<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0"
     xmlns:content="http://purl.org/rss/1.0/modules/content/"
     xmlns:dc="http://purl.org/dc/elements/1.1/"
     xmlns:dcterms="http://purl.org/dc/terms/"
     xmlns:media="http://search.yahoo.com/mrss/"
     xmlns:atom="http://www.w3.org/2005/Atom"
     xmlns:cf="https://www.futureplc.com/rss/content-flags"
>
    <channel>
                    <atom:link rel="alternate" hreflang="en-SG"
                       href="https://www.techradar.com/sg/feeds/tag/pro"
                       type="application/rss+xml"/>
                            <title><![CDATA[ Latest from TechRadar SG in Pro ]]></title>
                <link>https://www.techradar.com/sg/pro</link>
        <description><![CDATA[ All the latest pro content from the TechRadar  SG team ]]></description>
                                    <lastBuildDate>Tue, 17 Nov 2026 19:05:00 +0000</lastBuildDate>
                            <language>en</language>
                                <item>
                                                            <title><![CDATA[ Don't get irrit-AI-ted — new report claims a third of US chatbot users have sworn at or insulted their LLMs ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Anthropic has introduced new policies to prevent abuse of its Claude chatbot</strong></li><li><strong>Bad language directed at chatbots is on the rise, according to research from Searchable</strong></li><li><strong>Chatbots tend to resolve their errors more when users refrain from insulting them, according to the study</strong></li></ul><p>If you’ve ever got angry with your preferred AI chatbot, it could be worth rephrasing your angry response to the wrong answer. A report into bad behavior directed against AI tools has found that complaining with bad language or insults is less effective than querying the response with a straightforward statement.</p><p>The survey from Searchable found a wholly unsurprising 33% of US-based users swear or insult their chatbot. Compiling the data from an opt-in survey of 8 million messages exchanged with Claude, Gemini, and ChatGPT over 12 months, Searchable also found that residents of Nevada were more likely to insult their AI bots.</p><p>Following news of the survey, Anthropic has introduced a new set of usage policies to block cruelty in chats with Claude.</p><h2 id="tolerable-behavior">Tolerable behavior</h2><p>Monitoring of the messaging with chatbots ran between September 2025 and September 2026, with anonymized data and results grouped by age and location. The initial results found that for every 10,000 words in an exchange with an AI chatbot, an average of 19.1 were found to be derogatory. This was where Nevada came out on top, with its score of 22.7 rude or offensive words.</p><p>In contrast, New Jersey proved to be the most polite state, with 11.4 words. As far as cities go, what happens in Las Vegas doesn’t necessarily stay there when it comes to bad treatment of AI chatbots. There, 22 swear words or slurs per 10,000 were employed, while New York’s 12.3 words per 10,000 made it the politest city in the US.</p><p>The <a href="https://www.searchable.com/data/how-mean-are-americans-to-ai-chatbots" target="_blank">report</a> also highlighted a difference between men and women, age groups, and education levels.</p><h2 id="be-civil-to-the-ai">Be civil to the AI</h2><p>Of course, few people enter into a chatbot discussion with the express intention of causing offense to an LLM (if that was even possible). Rather, the misuse of language and casting aside of civility in this discussions seems to be caused by unsatisfactory results from the chatbots.</p><p>So, did the bad language have any impact? Well, it seems not.</p><p>By following rude exchanges to the end of the conversation, Searchable found that rudeness made the AI chatbot less likely to respond with a desirable result. </p><p>“By the end of the conversation, 31% of sweary complaints had been put right,” Content Lead Connor Lahey and Digital PR Lead Sean Fitzsimons wrote, “against 41% of plain corrections, and 35% were still not fixed, against 24%. When a complaint did get fixed, it took about the same number of replies either way.”</p><p>The solution, then, is to be nice to your AI chatbot, even when it gets things wrong.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/dont-get-irrit-ai-ted-new-report-claims-a-third-of-us-chatbot-users-have-sworn-at-or-insulted-their-llms</link>
                                                                            <description>
                            <![CDATA[ Users appear to be growing more frustrated with AI chatbots failing to follow instructions – now Anthropic is taking steps to stop it. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">TPZAaAFY3YiLBt5NMjqSbD</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/2yHEj7RR9QpUSvzcPym7Vf-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 19:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                                                                                    <dc:creator><![CDATA[ Christian Cawley ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/zBDYnjPnB2XPvhKbYX9Kuc-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Christian Cawley has extensive experience as a writer and editor in consumer electronics, IT and entertainment media. He has contributed to TechRadar since 2017 and has been published in Computer Weekly, Linux Format, ComputerActive, and other publications. &lt;/p&gt;&lt;p&gt;Beyond TechRadar, he heads up the team at smart home website Matter Alpha, and writes about retro gaming at Gaming Retro. &lt;/p&gt;&lt;p&gt;Formerly the editor responsible for Linux, Security, Programming, and DIY at MakeUseOf, Christian previously worked as a desktop and software support specialist in the public and private sectors.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/2yHEj7RR9QpUSvzcPym7Vf-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Angry man shouting at mobile phone while sitting at a desk]]></media:description>                                                            <media:text><![CDATA[Angry man shouting at mobile phone while sitting at a desk]]></media:text>
                                <media:title type="plain"><![CDATA[Angry man shouting at mobile phone while sitting at a desk]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/2yHEj7RR9QpUSvzcPym7Vf-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Anthropic has introduced new policies to prevent abuse of its Claude chatbot</strong></li><li><strong>Bad language directed at chatbots is on the rise, according to research from Searchable</strong></li><li><strong>Chatbots tend to resolve their errors more when users refrain from insulting them, according to the study</strong></li></ul><p>If you’ve ever got angry with your preferred AI chatbot, it could be worth rephrasing your angry response to the wrong answer. A report into bad behavior directed against AI tools has found that complaining with bad language or insults is less effective than querying the response with a straightforward statement.</p><p>The survey from Searchable found a wholly unsurprising 33% of US-based users swear or insult their chatbot. Compiling the data from an opt-in survey of 8 million messages exchanged with Claude, Gemini, and ChatGPT over 12 months, Searchable also found that residents of Nevada were more likely to insult their AI bots.</p><p>Following news of the survey, Anthropic has introduced a new set of usage policies to block cruelty in chats with Claude.</p><h2 id="tolerable-behavior">Tolerable behavior</h2><p>Monitoring of the messaging with chatbots ran between September 2025 and September 2026, with anonymized data and results grouped by age and location. The initial results found that for every 10,000 words in an exchange with an AI chatbot, an average of 19.1 were found to be derogatory. This was where Nevada came out on top, with its score of 22.7 rude or offensive words.</p><p>In contrast, New Jersey proved to be the most polite state, with 11.4 words. As far as cities go, what happens in Las Vegas doesn’t necessarily stay there when it comes to bad treatment of AI chatbots. There, 22 swear words or slurs per 10,000 were employed, while New York’s 12.3 words per 10,000 made it the politest city in the US.</p><p>The <a href="https://www.searchable.com/data/how-mean-are-americans-to-ai-chatbots" target="_blank">report</a> also highlighted a difference between men and women, age groups, and education levels.</p><h2 id="be-civil-to-the-ai">Be civil to the AI</h2><p>Of course, few people enter into a chatbot discussion with the express intention of causing offense to an LLM (if that was even possible). Rather, the misuse of language and casting aside of civility in this discussions seems to be caused by unsatisfactory results from the chatbots.</p><p>So, did the bad language have any impact? Well, it seems not.</p><p>By following rude exchanges to the end of the conversation, Searchable found that rudeness made the AI chatbot less likely to respond with a desirable result. </p><p>“By the end of the conversation, 31% of sweary complaints had been put right,” Content Lead Connor Lahey and Digital PR Lead Sean Fitzsimons wrote, “against 41% of plain corrections, and 35% were still not fixed, against 24%. When a complaint did get fixed, it took about the same number of replies either way.”</p><p>The solution, then, is to be nice to your AI chatbot, even when it gets things wrong.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Massive Kik data breach exposes more than 150 million images — experts say popular messaging app may have leaked incredibly sensitive material ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Researchers found an exposed database containing more than 150 million Kik profile images, including around 75 million original photos</strong></li><li><strong>Exposed images were publicly accessible online and included content involving minors, prompting reports to child-safety authorities</strong></li><li><strong>Database was secured after disclosure, but questions remain about exposure duration and potential prior access</strong></li></ul><p>Kik, a popular American instant messaging app, was leaking profile pictures on the clearweb, putting even children and young adults in harm's way. To make matters worse, some of the images that were leaking were disturbing enough for the researcher who found them to report them to the US National Center for Missing &amp; Exploited Children.</p><p>Security researcher Jeremiah Fowler <a href="https://www.expressvpn.com/research/kik-data-exposed" target="_blank">disclosed</a> discovering an unencrypted, non-password-protected database available to anyone with a browser and an internet connection - as long as they knew where to look.</p><p>This time around, he found a database containing just over 150 million images uploaded between 2021 and 2026. They were marked as thumbnails and originals, meaning that at least half - some 75 million - were unique, large-format images and photos. Fowler later determined they were marked as “profiles” and associated with Kik.</p><h2 id="kik-locks-it-down">Kik locks it down</h2><p>Launched in 2010, Kik has historically reported more than 300 million registered accounts, although its current active-user figures have not been publicly disclosed. Google Play lists more than 100 million downloads, and Fowler claims it has more than 120 million registered users and around 15 million active users.</p><p>Kik is owned by MediaLab, a US-based company that also owns social networking platform Whisper and other digital media properties. </p><p>“The files were hosted in a <a href="https://www.techradar.com/best/best-cloud-storage" target="_blank">cloud storage database</a> and could also be viewed in any web browser through the direct hosting URL and a subfolder on the primary Kik domain,” Fowler said.</p><p>“The name of the database and the fact that the image resources were also hosted on a Kik-associated subdomain provided strong evidence that the exposed records were associated with the platform.”</p><p>The researcher notified Kik of his findings, and the database was quickly locked down. However, the company did not reply to a request for comment, so we don’t know how many people were exposed (75 million original pictures does not necessarily mean 75 million people - there are likely duplicates and fake accounts), for how long the database remained open, who managed it (Kik, MediaLab, or a third party), and if any malicious actors found it before Fowler did (which is always a possibility). </p><h2 id="a-dark-turn">A dark turn</h2><p>This is all relatively standard news in the world of unintentionally misconfigured databases. The findings took a dark turn when Fowler discovered that in a sample of some 2,000 images, there were 66 profile images appearing to contain child sexual abuse material (CSAM). </p><p>Fowler suggests Kik is primarily a platform for teens and young adults: “In 2016, the messaging platform was used by approximately 40% of American teenagers and a recent article published by the American Enterprise Institute stated that 70% of Kik users are in the 13 to 24 age bracket.</p><p>“In 2025, Kik's terms of service were updated to state that the app is only for individuals aged 18 and older. In my limited review of the publicly exposed images, I saw numerous images of individuals who appeared to be under the age of 18," the report notes.</p><p>While the report does plenty of hedging, saying that a single image is not enough to determine someone’s age, it does state that Fowler reported some of the images to the National Center for Missing &amp; Exploited Children (NCMEC) and the Connecticut Attorney General’s office.</p><p>“In a random selection from the limited sample of files, I saw a large number of profile images containing nudity or adult-themed content,” the researcher stressed.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/massive-kik-data-breach-exposes-more-than-150-million-images-experts-say-popular-messaging-app-may-have-leaked-incredibly-sensitive-material</link>
                                                                            <description>
                            <![CDATA[ Some of the images could be described as containing child sexual abuse material and were reported to authorities. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">tKKP27JG8rEuVLieCEQwVi</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/DyF5dEL5UeH88zBmpwo3EU-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 18:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/DyF5dEL5UeH88zBmpwo3EU-1920-80.jpg">
                                                            <media:credit><![CDATA[ExpressVPN/Jeremiah Fowler]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Kik]]></media:description>                                                            <media:text><![CDATA[Kik]]></media:text>
                                <media:title type="plain"><![CDATA[Kik]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/DyF5dEL5UeH88zBmpwo3EU-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Researchers found an exposed database containing more than 150 million Kik profile images, including around 75 million original photos</strong></li><li><strong>Exposed images were publicly accessible online and included content involving minors, prompting reports to child-safety authorities</strong></li><li><strong>Database was secured after disclosure, but questions remain about exposure duration and potential prior access</strong></li></ul><p>Kik, a popular American instant messaging app, was leaking profile pictures on the clearweb, putting even children and young adults in harm's way. To make matters worse, some of the images that were leaking were disturbing enough for the researcher who found them to report them to the US National Center for Missing &amp; Exploited Children.</p><p>Security researcher Jeremiah Fowler <a href="https://www.expressvpn.com/research/kik-data-exposed" target="_blank">disclosed</a> discovering an unencrypted, non-password-protected database available to anyone with a browser and an internet connection - as long as they knew where to look.</p><p>This time around, he found a database containing just over 150 million images uploaded between 2021 and 2026. They were marked as thumbnails and originals, meaning that at least half - some 75 million - were unique, large-format images and photos. Fowler later determined they were marked as “profiles” and associated with Kik.</p><h2 id="kik-locks-it-down">Kik locks it down</h2><p>Launched in 2010, Kik has historically reported more than 300 million registered accounts, although its current active-user figures have not been publicly disclosed. Google Play lists more than 100 million downloads, and Fowler claims it has more than 120 million registered users and around 15 million active users.</p><p>Kik is owned by MediaLab, a US-based company that also owns social networking platform Whisper and other digital media properties. </p><p>“The files were hosted in a <a href="https://www.techradar.com/best/best-cloud-storage" target="_blank">cloud storage database</a> and could also be viewed in any web browser through the direct hosting URL and a subfolder on the primary Kik domain,” Fowler said.</p><p>“The name of the database and the fact that the image resources were also hosted on a Kik-associated subdomain provided strong evidence that the exposed records were associated with the platform.”</p><p>The researcher notified Kik of his findings, and the database was quickly locked down. However, the company did not reply to a request for comment, so we don’t know how many people were exposed (75 million original pictures does not necessarily mean 75 million people - there are likely duplicates and fake accounts), for how long the database remained open, who managed it (Kik, MediaLab, or a third party), and if any malicious actors found it before Fowler did (which is always a possibility). </p><h2 id="a-dark-turn">A dark turn</h2><p>This is all relatively standard news in the world of unintentionally misconfigured databases. The findings took a dark turn when Fowler discovered that in a sample of some 2,000 images, there were 66 profile images appearing to contain child sexual abuse material (CSAM). </p><p>Fowler suggests Kik is primarily a platform for teens and young adults: “In 2016, the messaging platform was used by approximately 40% of American teenagers and a recent article published by the American Enterprise Institute stated that 70% of Kik users are in the 13 to 24 age bracket.</p><p>“In 2025, Kik's terms of service were updated to state that the app is only for individuals aged 18 and older. In my limited review of the publicly exposed images, I saw numerous images of individuals who appeared to be under the age of 18," the report notes.</p><p>While the report does plenty of hedging, saying that a single image is not enough to determine someone’s age, it does state that Fowler reported some of the images to the National Center for Missing &amp; Exploited Children (NCMEC) and the Connecticut Attorney General’s office.</p><p>“In a random selection from the limited sample of files, I saw a large number of profile images containing nudity or adult-themed content,” the researcher stressed.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Some of the world's largest brands hit following Optimizely flaw — Salesforce, CNN, Fox News, BBC sites hit with malware ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Attackers compromised Optimizely and injected malicious JavaScript into more than 100 popular websites through a trusted content delivery path</strong></li><li><strong>The payload stole cryptocurrency, harvested payment-card data, captured user information, and targeted iPhone visitors with exploit kits</strong></li><li><strong>Optimizely removed the malicious code, revoked access, and patched the signup-flow vulnerability used in the attack</strong></li></ul><p>Threat actors broke into Optimizely and used it to deploy malicious code to more than 100 popular websites across the internet, with the code replacing people’s cryptocurrency wallet addresses, stealing credit card information, and even allowing attackers to take over iPhones, experts have said.</p><p>Optimizely is a digital experience platform that businesses use to manage their websites, create and optimize digital content, run A/B tests, and personalize customer experiences. It is used primarily by marketing and product teams to experiment with different features of the website, improving engagement, conversions, and ultimately - sales. </p><p>It is a hugely popular tool, one that’s used by hundreds of organizations in all sorts of industries. The attacker deployed 188 snippets onto 178 organizations. Most of them were in banking &amp; finance (32), technology &amp; SaaS (23), and retail &amp; fashion (20), with some major names on the list: HSBC US, Equifax, Zoom, Salesforce, IKEA, ASOS, Peacock, CNN, BBC, Fox News, Alaska Airlines / Hawaiian Airlines, Virgin Media, Virgin Media O2, Vodafone Australia, O2, Ledger, and many others.</p><h2 id="exploiting-a-flaw">Exploiting a flaw</h2><p>In a <a href="https://blog.report-uri.com/worlds-largest-brands-infected-with-malware-via-optimizely/" target="_blank" rel="nofollow">blog post</a>, security researcher Scott Helme, who discovered the attack, explains what happened - each customer loads a JavaScript snippet from Optimizely’s Content Delivery Network (CDN), containing the company’s code, the customer’s experiments, and an optional field called Project JavaScript. This is custom code the customer writes in their own Optimizely dashboard which runs on every page. </p><p>According to Helme, the attacker wrote their own code into the Project JavaScript of customer projects and the file was served to multiple victims. It still went out from cdn.optimizely.com, with Optimizely’s certificate, from a known and trusted host. </p><p>In a statement, Optimizely said the attacker exploited a flaw in its Experimentation account signup flow to inject the malicious JavaScript code.</p><p>"Earlier today, Optimizely identified a security incident in which an unauthorized party exploited a vulnerability in the Experimentation account signup flow," Optimizely said.</p><p>"They used it to add malicious JavaScript to some customer projects. We have blocked access to the vulnerable path and deployed a fix. We have removed the injected code from affected projects. We have also ended active sessions, revoked tokens, and reset access for affected user accounts. Finally, we have blocked any code from being delivered in Cloudflare.</p><p>Helme says the threat actors rebuilt the payload 26 times in one day, adding features and cryptocurrency wallets. The payload did four key things: stole people’s crypto, skimmed compromised websites for customer credit cards, stole cookies and took screenshots, and deployed iOS hacking toolkits.</p><h2 id="fighting-back">Fighting back</h2><p>Crypto theft worked through clipboard jacking: the code would monitor the clipboard for copied wallets and replace them with attacker-owned ones when pasting. </p><p>That way, victims would copy one address and unknowingly paste an entirely different one, sending money to the attackers. The skimmer would monitor the sites for card numbers, expiry dates, and CVV secrets, and when it finds all three, it would send them to an attacker-owned server. Cards entered in hosted payment iframes (Stripe, Adyen and similar) were out of its reach, it was said. </p><p>Victims opening the websites on iPhones and iPads were served Coruna and DarkSword, exploit kits designed to abuse unpatched iOS vulnerabilities and steal sensitive information, including cryptocurrency wallet credentials and recovery phrases.</p><p>Helme says the threat actors stole just over $7,000 in this campaign. Most of it was in bitcoin (around $5,300), with the rest split between litecoin, USDT, and ether. </p><p>Most of the malicious code has now been removed, Helme stressed. Less than 24 hours after the first attack was spotted, the number of infected snippets was down to just one - TalkTalk. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/some-of-the-worlds-largest-brands-hit-following-optimizely-flaw-salesforce-cnn-fox-news-bbc-sites-hit-with-malware</link>
                                                                            <description>
                            <![CDATA[ A bug in Optimizely allowed an attacker to deploy malicious JavaScript across the web, compromising popular retail and tech sites. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">DnsMs6XAMLAnA3Tb84aQ3G</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Ri2dNNTvgmKGsMuhNDCavZ-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 17:15:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Ri2dNNTvgmKGsMuhNDCavZ-1920-80.jpg">
                                                            <media:credit><![CDATA[Adobe]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Dark web monitoring]]></media:description>                                                            <media:text><![CDATA[Dark web monitoring]]></media:text>
                                <media:title type="plain"><![CDATA[Dark web monitoring]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Ri2dNNTvgmKGsMuhNDCavZ-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Attackers compromised Optimizely and injected malicious JavaScript into more than 100 popular websites through a trusted content delivery path</strong></li><li><strong>The payload stole cryptocurrency, harvested payment-card data, captured user information, and targeted iPhone visitors with exploit kits</strong></li><li><strong>Optimizely removed the malicious code, revoked access, and patched the signup-flow vulnerability used in the attack</strong></li></ul><p>Threat actors broke into Optimizely and used it to deploy malicious code to more than 100 popular websites across the internet, with the code replacing people’s cryptocurrency wallet addresses, stealing credit card information, and even allowing attackers to take over iPhones, experts have said.</p><p>Optimizely is a digital experience platform that businesses use to manage their websites, create and optimize digital content, run A/B tests, and personalize customer experiences. It is used primarily by marketing and product teams to experiment with different features of the website, improving engagement, conversions, and ultimately - sales. </p><p>It is a hugely popular tool, one that’s used by hundreds of organizations in all sorts of industries. The attacker deployed 188 snippets onto 178 organizations. Most of them were in banking &amp; finance (32), technology &amp; SaaS (23), and retail &amp; fashion (20), with some major names on the list: HSBC US, Equifax, Zoom, Salesforce, IKEA, ASOS, Peacock, CNN, BBC, Fox News, Alaska Airlines / Hawaiian Airlines, Virgin Media, Virgin Media O2, Vodafone Australia, O2, Ledger, and many others.</p><h2 id="exploiting-a-flaw">Exploiting a flaw</h2><p>In a <a href="https://blog.report-uri.com/worlds-largest-brands-infected-with-malware-via-optimizely/" target="_blank" rel="nofollow">blog post</a>, security researcher Scott Helme, who discovered the attack, explains what happened - each customer loads a JavaScript snippet from Optimizely’s Content Delivery Network (CDN), containing the company’s code, the customer’s experiments, and an optional field called Project JavaScript. This is custom code the customer writes in their own Optimizely dashboard which runs on every page. </p><p>According to Helme, the attacker wrote their own code into the Project JavaScript of customer projects and the file was served to multiple victims. It still went out from cdn.optimizely.com, with Optimizely’s certificate, from a known and trusted host. </p><p>In a statement, Optimizely said the attacker exploited a flaw in its Experimentation account signup flow to inject the malicious JavaScript code.</p><p>"Earlier today, Optimizely identified a security incident in which an unauthorized party exploited a vulnerability in the Experimentation account signup flow," Optimizely said.</p><p>"They used it to add malicious JavaScript to some customer projects. We have blocked access to the vulnerable path and deployed a fix. We have removed the injected code from affected projects. We have also ended active sessions, revoked tokens, and reset access for affected user accounts. Finally, we have blocked any code from being delivered in Cloudflare.</p><p>Helme says the threat actors rebuilt the payload 26 times in one day, adding features and cryptocurrency wallets. The payload did four key things: stole people’s crypto, skimmed compromised websites for customer credit cards, stole cookies and took screenshots, and deployed iOS hacking toolkits.</p><h2 id="fighting-back">Fighting back</h2><p>Crypto theft worked through clipboard jacking: the code would monitor the clipboard for copied wallets and replace them with attacker-owned ones when pasting. </p><p>That way, victims would copy one address and unknowingly paste an entirely different one, sending money to the attackers. The skimmer would monitor the sites for card numbers, expiry dates, and CVV secrets, and when it finds all three, it would send them to an attacker-owned server. Cards entered in hosted payment iframes (Stripe, Adyen and similar) were out of its reach, it was said. </p><p>Victims opening the websites on iPhones and iPads were served Coruna and DarkSword, exploit kits designed to abuse unpatched iOS vulnerabilities and steal sensitive information, including cryptocurrency wallet credentials and recovery phrases.</p><p>Helme says the threat actors stole just over $7,000 in this campaign. Most of it was in bitcoin (around $5,300), with the rest split between litecoin, USDT, and ether. </p><p>Most of the malicious code has now been removed, Helme stressed. Less than 24 hours after the first attack was spotted, the number of infected snippets was down to just one - TalkTalk. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Infoblox confirms company and customer data exposed in SharePoint breach ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Infoblox is investigating a supply-chain breach in which attackers allegedly used a vendor’s access to steal employee and customer data</strong></li><li><strong>Exposed information reportedly includes employee personal details and customer network configuration backups</strong></li><li><strong>Company is offering identity protection services and warning affected individuals about phishing and fraud risks</strong></li></ul><p>Infoblox has said it is currently investigating reports of a cyberattack and sensitive data theft. </p><p>Initial reports describe it as a third-party supply chain attack in which employee data and customer network information were exfiltrated.</p><p><em></em><a href="https://www.theregister.com/security/2026/11/16/infoblox-customer-and-staff-data-exposed-in-sharepoint-breach/5310280" target="_blank"><em>The Register</em></a> revealed how Infoblox worked with a third-party data migration vendor which was granted trusted access to its SharePoint system, meaning the unidentified attackers apparently obtained login credentials for vendor accounts, used them to access Infoblox’s Microsoft SharePoint instances, and pulled data straight from there.</p><h2 id="nios-deployments-and-personal-data">NIOS deployments and personal data</h2><p>Exactly who the attackers are, or how they obtained the credentials, was not explained. Usually, they either steal the passwords with infostealers, or trick victims into handing over the credentials themselves. </p><p>The former method is easy in theory - get the victim to download malware and the program sends over stored passwords, authentication cookies, sensitive documents, and other files. The latter is somewhat more difficult since it involves creating a fake login page of a known - and used - online service that secretly relays all submitted information, sending it to the victim, and hoping they try to log in.</p><p>This is usually done through phishing emails that warn the victim they will lose access to accounts, or that the affiliated data will be deleted, making them hurry and not pay attention to what they’re doing. </p><p>Among the nabbed information are some customers’ Network Identity Operating System (NIOS) deployments, <em>The Register</em> claims. “These are saved snapshots of customers’ network data, such as DNS, DHCP configurations, Grid databases, and reporting logs, that can be used to fully recover the appliance or grid in case of a system failure or faulty upgrade.”</p><p>There’s plenty of value in stolen NIOS config backups. These can serve as blueprints of a corporate network, including DNS infrastructure, DHCP settings, internal IP address range, and connected systems. </p><p>This kind of information can be useful to map a network, identify critical infrastructure, locate potential weak points, and plan targeted attacks. DNS configurations, for example, can reveal internal hostnames or domain structures, helping crooks identify valuable targets or craft custom-tailored and highly convincing phishing lures. Access to Grid databases and recovery snapshots could also expose sensitive operational details that make it easier to impersonate legitimate infrastructure or disrupt network services. </p><p>The attackers also apparently stole employee data. Citing an email Infoblox CEO Scott Harrell allegedly sent to employees, <em>The Register</em> says the attackers took financial information such as direct deposit details, contact info, dates of birth, and Social Security numbers for employees and dependents. </p><h2 id="offering-help">Offering help</h2><p>This is more than enough information to launch phishing attacks which could very easily turn into ransomware or wire fraud. </p><p>When asked for comment, the company told <em>The Register</em> that it was “investigating a security event involving unauthorized access to certain company information through a compromised third-party data migration vendor.” No names or concrete figures anywhere in the statement. </p><p>“We have sent notifications about this security event and are providing support and guidance as appropriate,” the Infoblox spokesperson said. “Based on our investigation to date, we have no indication that any Infoblox products, cloud services, software updates or build-and-release processes were involved. Our priority is to fully investigate and resolve this issue."</p><p>The company is now offering money to the victims - $100 of “spot bonus” to help individuals purchase password management services, as well as two years of free identity theft and credit monitoring services - these days a standard practice in similar incidents. Infoblox also urged everyone to be "especially alert” for phishing, vishing, fraud, or other suspicious communications.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/infoblox-confirms-company-and-customer-data-exposed-in-sharepoint-breach</link>
                                                                            <description>
                            <![CDATA[ Infoblox says its investigation is ongoing, but it suspects a third-party supply chain attack. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">346FH7wkNMxFdmrac7PJCS</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/37uyEphcLreEFNUVCQzurn-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 16:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/37uyEphcLreEFNUVCQzurn-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[World Password Day 2025]]></media:description>                                                            <media:text><![CDATA[World Password Day 2025]]></media:text>
                                <media:title type="plain"><![CDATA[World Password Day 2025]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/37uyEphcLreEFNUVCQzurn-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Infoblox is investigating a supply-chain breach in which attackers allegedly used a vendor’s access to steal employee and customer data</strong></li><li><strong>Exposed information reportedly includes employee personal details and customer network configuration backups</strong></li><li><strong>Company is offering identity protection services and warning affected individuals about phishing and fraud risks</strong></li></ul><p>Infoblox has said it is currently investigating reports of a cyberattack and sensitive data theft. </p><p>Initial reports describe it as a third-party supply chain attack in which employee data and customer network information were exfiltrated.</p><p><em></em><a href="https://www.theregister.com/security/2026/11/16/infoblox-customer-and-staff-data-exposed-in-sharepoint-breach/5310280" target="_blank"><em>The Register</em></a> revealed how Infoblox worked with a third-party data migration vendor which was granted trusted access to its SharePoint system, meaning the unidentified attackers apparently obtained login credentials for vendor accounts, used them to access Infoblox’s Microsoft SharePoint instances, and pulled data straight from there.</p><h2 id="nios-deployments-and-personal-data">NIOS deployments and personal data</h2><p>Exactly who the attackers are, or how they obtained the credentials, was not explained. Usually, they either steal the passwords with infostealers, or trick victims into handing over the credentials themselves. </p><p>The former method is easy in theory - get the victim to download malware and the program sends over stored passwords, authentication cookies, sensitive documents, and other files. The latter is somewhat more difficult since it involves creating a fake login page of a known - and used - online service that secretly relays all submitted information, sending it to the victim, and hoping they try to log in.</p><p>This is usually done through phishing emails that warn the victim they will lose access to accounts, or that the affiliated data will be deleted, making them hurry and not pay attention to what they’re doing. </p><p>Among the nabbed information are some customers’ Network Identity Operating System (NIOS) deployments, <em>The Register</em> claims. “These are saved snapshots of customers’ network data, such as DNS, DHCP configurations, Grid databases, and reporting logs, that can be used to fully recover the appliance or grid in case of a system failure or faulty upgrade.”</p><p>There’s plenty of value in stolen NIOS config backups. These can serve as blueprints of a corporate network, including DNS infrastructure, DHCP settings, internal IP address range, and connected systems. </p><p>This kind of information can be useful to map a network, identify critical infrastructure, locate potential weak points, and plan targeted attacks. DNS configurations, for example, can reveal internal hostnames or domain structures, helping crooks identify valuable targets or craft custom-tailored and highly convincing phishing lures. Access to Grid databases and recovery snapshots could also expose sensitive operational details that make it easier to impersonate legitimate infrastructure or disrupt network services. </p><p>The attackers also apparently stole employee data. Citing an email Infoblox CEO Scott Harrell allegedly sent to employees, <em>The Register</em> says the attackers took financial information such as direct deposit details, contact info, dates of birth, and Social Security numbers for employees and dependents. </p><h2 id="offering-help">Offering help</h2><p>This is more than enough information to launch phishing attacks which could very easily turn into ransomware or wire fraud. </p><p>When asked for comment, the company told <em>The Register</em> that it was “investigating a security event involving unauthorized access to certain company information through a compromised third-party data migration vendor.” No names or concrete figures anywhere in the statement. </p><p>“We have sent notifications about this security event and are providing support and guidance as appropriate,” the Infoblox spokesperson said. “Based on our investigation to date, we have no indication that any Infoblox products, cloud services, software updates or build-and-release processes were involved. Our priority is to fully investigate and resolve this issue."</p><p>The company is now offering money to the victims - $100 of “spot bonus” to help individuals purchase password management services, as well as two years of free identity theft and credit monitoring services - these days a standard practice in similar incidents. Infoblox also urged everyone to be "especially alert” for phishing, vishing, fraud, or other suspicious communications.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Why adaptability beats capacity for data centers ]]></title>
                                                                                                <dc:content><![CDATA[ <p>There’s always a quiet moment of calm in my morning that usually comes just after the 6 a.m. handover from the night team, and before meetings with customers and prospects get started. It’s time I reserve for reflecting and planning, looking at both the week ahead and working on the longer-term requirements our facility needs to meet.</p><p>It’s the latter of those priorities that’s taking more of my time these days, and it reflects the changing nature of the conversations I’m having with CIOs and technology leaders. Two years ago, most customers, especially those in regulated sectors such as finance and healthcare, were asking about uptime, certification and cost. </p><p>Today, the first question is almost always about <a href="https://www.techradar.com/best/best-ai-tools">AI</a>. </p><p>They want to know if the facility can support the capacity they’re going to need, and whether it can do so on a timeline that matches how quickly their workloads are scaling. These are fair and important questions, and, in all honesty, they are not always easy to answer. </p><p>The increase in demand, driven by AI, means decisions we previously had years to make, now need to be made in a matter of months. Not impossible, but it fundamentally changes the way we think about designing and running facilities longer term. </p><h2 id="density-not-just-capacity-is-what-matters-for-ai-workloads">Density, not just capacity, is what matters for AI workloads</h2><p>Around the world, capacity challenges dominate headlines, but the <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> story underneath it is almost entirely about power density, cooling, and grid capacity. Historically, infrastructure was designed for low power density, whereas today, AI needs much more power per rack as well as more advanced cooling. </p><p>A good example is a rack that housed standard enterprise workloads at 5 to 10 kVA two years ago is now being asked to support 15 kVA within the next 18 months, and we’re already planning to support 25 to 30 kVA within the next few years. That is not an incremental change. It requires a different physical design that provides more density in an efficient way, often through modular Performance Optimized Data Centre (POD) <a href="https://www.techradar.com/best/best-architecture-software">architecture</a> and in-row cooling. This allows us to add power and cooling gradually, controlling investment, improving efficiency, and ultimately, allowing us to increase capacity as demand grows.</p><p>The lack of density for AI workloads is why many established markets, including some of Europe's <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> hubs, are struggling to add capacity at the pace required by demand. </p><p>For CIOs evaluating a provider, the useful question isn't “do you have any capacity today?” it's “how does your design accommodate the density I'll need in three years?” because a facility built for yesterday's power needs can quickly become an operational risk if it can't be reconfigured without a lengthy rebuild.</p><p>A good example of this is a POD we recently reconfigured within 30 days to meet a customer’s changing requirements, changing 40 racks at 5 kVA, to 20 racks at 10 kVA to support high-density demands. It was a significant project, and the kind of turnaround that's only possible when a facility is designed for change and adaptability, rather than attempting to bolt flexibility on afterwards.</p><h2 id="cooling-design-has-to-keep-pace-with-density">Cooling design has to keep pace with density</h2><p>Managing excess heat caused by the increased processing demands of AI workloads is becoming an increasingly important part of infrastructure <a href="https://www.techradar.com/best/best-graphic-design-software">design</a> and customer consideration. </p><p>The good news is that cooling technology is improving all the time, making high density possible without needing to prioritize colder climates. </p><p>We mostly use air cooling with in-row technology, and because our design is modular, we can increase cooling as the IT load grows. For example, our customers typically join with an initial setup and gradually scale their capacity as their requirements grow. </p><p>As the customer’s IT load increases, we progressively activate the in-row cooling units, ensuring that cooling capacity is only deployed when and where it is needed. We also use a refrigerant-based cooling system rather than a water-based one, which keeps our water consumption low to begin with </p><p>This approach allows us to operate more efficiently when it comes to both energy and water use, which, as we know, is under increasing scrutiny from regulators and communities alike, and as a limited resource, needs to be used responsibly.</p><h2 id="local-footprint-must-be-part-of-the-design-conversation-too">Local footprint must be part of the design conversation, too</h2><p>It’s been hard to escape the recent tranche of negative media coverage about data centers taking valuable resources (land, water and power) from local communities. It’s true that they have an impact on the local community, but it’s unfair to frame this as entirely negative. </p><p>In my experience, a site's economic footprint is much larger than the building itself. Beyond the direct operational jobs, data centers support a much wider chain of indirect employment in sectors such as security, maintenance, cleaning, logistics, and transport. They also drive investment in the surrounding energy and connectivity infrastructure.</p><p>The ability to serve the local community and economy, as well as our customers, is an integral part of my role. It’s also something I encourage technology leaders to consider as a legitimate part of their due diligence. After all, banks, government services, <a href="https://www.techradar.com/best/best-cloud-computing-services">cloud providers</a>, and now AI, all depend on the same underlying capacity, and the locations that can offer credible, well-designed and well-run infrastructure, also attract more technology investment, digital skills, and the surrounding ecosystem that a growing digital economy needs. </p><h2 id="sustainability-has-to-be-operational-not-aspirational">Sustainability has to be operational, not aspirational</h2><p>The scrutiny on data centers' environmental footprint is only intensifying as AI-driven demand grows, and rightly so. When it comes to managing a facility, this means building sustainability into daily operations, rather than treating it as a future aspiration or initiative that is bolted on after the fact.  </p><p>In practice, this means ensuring efficient cooling design, transparent community engagement, and genuine investment in the skilled engineering talent needed to run higher-density infrastructure safely. At Ilkari, we do this through our POD-based design and in-row cooling architecture, which allow us to scale cooling capacity efficiently as IT load grows, and maintain efficient and sustainable data center operations.</p><p>It also means working with local governments to make sure the right investments are being made in local power and energy infrastructure to allow us to operate in a way that is sustainable for the local community.  </p><h2 id="the-infrastructure-design-questions-cios-should-be-asking">The infrastructure design questions CIOs should be asking</h2><p>The reality of running a data centre as AI demand increases means a fundamental shift in how facilities need to be designed and built. </p><p>For CIOs thinking about what infrastructure requirements their AI workloads need, my advice is not to start with a capacity <a href="https://www.techradar.com/best/spreadsheet-software">spreadsheet</a>, but with a short list of broader questions, including: </p><p>- How does this provider's design scale with density, not just floor space? </p><p>- Can it demonstrate its ability to meet your requirements under real conditions, not just on paper? </p><p>- Is sustainability built into day-to-day operations (cooling design, community engagement, skilled engineering talent), or is it a separate initiative?</p><p>- As density increases, has the provider planned for the point where air cooling is no longer sufficient, and what alternative cooling technologies and infrastructure would that require?</p><p>In my experience, the organizations getting this right aren't the ones with the flashiest AI roadmap, they are the ones designing and building for change from the outset.</p><p><em></em><a href="https://www.techradar.com/news/best-dedicated-server-hosting-providers"><em>We've reviewed and ranked the best dedicated hosting services.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/why-adaptability-beats-capacity-for-data-centers</link>
                                                                            <description>
                            <![CDATA[ AI is reshaping data center design, density, cooling and sustainability ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">xtKuPeNmYbQheMRkMjZemZ</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/EXMLBYo5k7EwcuyYg9vmmM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 15:28:21 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Eduardo Espinel ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/EXMLBYo5k7EwcuyYg9vmmM-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                        <media:description><![CDATA[As AI demand accelerates, data centers must adapt faster, smarter and more sustainably.]]></media:description>                                                            <media:text><![CDATA[A long corridor with a sleek black floor, glowing green lights in the ceiling and rows of LEDS on either wall]]></media:text>
                                <media:title type="plain"><![CDATA[A long corridor with a sleek black floor, glowing green lights in the ceiling and rows of LEDS on either wall]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/EXMLBYo5k7EwcuyYg9vmmM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>There’s always a quiet moment of calm in my morning that usually comes just after the 6 a.m. handover from the night team, and before meetings with customers and prospects get started. It’s time I reserve for reflecting and planning, looking at both the week ahead and working on the longer-term requirements our facility needs to meet.</p><p>It’s the latter of those priorities that’s taking more of my time these days, and it reflects the changing nature of the conversations I’m having with CIOs and technology leaders. Two years ago, most customers, especially those in regulated sectors such as finance and healthcare, were asking about uptime, certification and cost. </p><p>Today, the first question is almost always about <a href="https://www.techradar.com/best/best-ai-tools">AI</a>. </p><p>They want to know if the facility can support the capacity they’re going to need, and whether it can do so on a timeline that matches how quickly their workloads are scaling. These are fair and important questions, and, in all honesty, they are not always easy to answer. </p><p>The increase in demand, driven by AI, means decisions we previously had years to make, now need to be made in a matter of months. Not impossible, but it fundamentally changes the way we think about designing and running facilities longer term. </p><h2 id="density-not-just-capacity-is-what-matters-for-ai-workloads">Density, not just capacity, is what matters for AI workloads</h2><p>Around the world, capacity challenges dominate headlines, but the <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> story underneath it is almost entirely about power density, cooling, and grid capacity. Historically, infrastructure was designed for low power density, whereas today, AI needs much more power per rack as well as more advanced cooling. </p><p>A good example is a rack that housed standard enterprise workloads at 5 to 10 kVA two years ago is now being asked to support 15 kVA within the next 18 months, and we’re already planning to support 25 to 30 kVA within the next few years. That is not an incremental change. It requires a different physical design that provides more density in an efficient way, often through modular Performance Optimized Data Centre (POD) <a href="https://www.techradar.com/best/best-architecture-software">architecture</a> and in-row cooling. This allows us to add power and cooling gradually, controlling investment, improving efficiency, and ultimately, allowing us to increase capacity as demand grows.</p><p>The lack of density for AI workloads is why many established markets, including some of Europe's <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> hubs, are struggling to add capacity at the pace required by demand. </p><p>For CIOs evaluating a provider, the useful question isn't “do you have any capacity today?” it's “how does your design accommodate the density I'll need in three years?” because a facility built for yesterday's power needs can quickly become an operational risk if it can't be reconfigured without a lengthy rebuild.</p><p>A good example of this is a POD we recently reconfigured within 30 days to meet a customer’s changing requirements, changing 40 racks at 5 kVA, to 20 racks at 10 kVA to support high-density demands. It was a significant project, and the kind of turnaround that's only possible when a facility is designed for change and adaptability, rather than attempting to bolt flexibility on afterwards.</p><h2 id="cooling-design-has-to-keep-pace-with-density">Cooling design has to keep pace with density</h2><p>Managing excess heat caused by the increased processing demands of AI workloads is becoming an increasingly important part of infrastructure <a href="https://www.techradar.com/best/best-graphic-design-software">design</a> and customer consideration. </p><p>The good news is that cooling technology is improving all the time, making high density possible without needing to prioritize colder climates. </p><p>We mostly use air cooling with in-row technology, and because our design is modular, we can increase cooling as the IT load grows. For example, our customers typically join with an initial setup and gradually scale their capacity as their requirements grow. </p><p>As the customer’s IT load increases, we progressively activate the in-row cooling units, ensuring that cooling capacity is only deployed when and where it is needed. We also use a refrigerant-based cooling system rather than a water-based one, which keeps our water consumption low to begin with </p><p>This approach allows us to operate more efficiently when it comes to both energy and water use, which, as we know, is under increasing scrutiny from regulators and communities alike, and as a limited resource, needs to be used responsibly.</p><h2 id="local-footprint-must-be-part-of-the-design-conversation-too">Local footprint must be part of the design conversation, too</h2><p>It’s been hard to escape the recent tranche of negative media coverage about data centers taking valuable resources (land, water and power) from local communities. It’s true that they have an impact on the local community, but it’s unfair to frame this as entirely negative. </p><p>In my experience, a site's economic footprint is much larger than the building itself. Beyond the direct operational jobs, data centers support a much wider chain of indirect employment in sectors such as security, maintenance, cleaning, logistics, and transport. They also drive investment in the surrounding energy and connectivity infrastructure.</p><p>The ability to serve the local community and economy, as well as our customers, is an integral part of my role. It’s also something I encourage technology leaders to consider as a legitimate part of their due diligence. After all, banks, government services, <a href="https://www.techradar.com/best/best-cloud-computing-services">cloud providers</a>, and now AI, all depend on the same underlying capacity, and the locations that can offer credible, well-designed and well-run infrastructure, also attract more technology investment, digital skills, and the surrounding ecosystem that a growing digital economy needs. </p><h2 id="sustainability-has-to-be-operational-not-aspirational">Sustainability has to be operational, not aspirational</h2><p>The scrutiny on data centers' environmental footprint is only intensifying as AI-driven demand grows, and rightly so. When it comes to managing a facility, this means building sustainability into daily operations, rather than treating it as a future aspiration or initiative that is bolted on after the fact.  </p><p>In practice, this means ensuring efficient cooling design, transparent community engagement, and genuine investment in the skilled engineering talent needed to run higher-density infrastructure safely. At Ilkari, we do this through our POD-based design and in-row cooling architecture, which allow us to scale cooling capacity efficiently as IT load grows, and maintain efficient and sustainable data center operations.</p><p>It also means working with local governments to make sure the right investments are being made in local power and energy infrastructure to allow us to operate in a way that is sustainable for the local community.  </p><h2 id="the-infrastructure-design-questions-cios-should-be-asking">The infrastructure design questions CIOs should be asking</h2><p>The reality of running a data centre as AI demand increases means a fundamental shift in how facilities need to be designed and built. </p><p>For CIOs thinking about what infrastructure requirements their AI workloads need, my advice is not to start with a capacity <a href="https://www.techradar.com/best/spreadsheet-software">spreadsheet</a>, but with a short list of broader questions, including: </p><p>- How does this provider's design scale with density, not just floor space? </p><p>- Can it demonstrate its ability to meet your requirements under real conditions, not just on paper? </p><p>- Is sustainability built into day-to-day operations (cooling design, community engagement, skilled engineering talent), or is it a separate initiative?</p><p>- As density increases, has the provider planned for the point where air cooling is no longer sufficient, and what alternative cooling technologies and infrastructure would that require?</p><p>In my experience, the organizations getting this right aren't the ones with the flashiest AI roadmap, they are the ones designing and building for change from the outset.</p><p><em></em><a href="https://www.techradar.com/news/best-dedicated-server-hosting-providers"><em>We've reviewed and ranked the best dedicated hosting services.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Half of businesses run workloads in the wrong place — mainly because they can't be bothered to change their ways ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Unsuitable cloud infrastructure is adding latency and reducing performance</strong></li><li><strong>Enterprises worry they'd have to re-architecture applications just to move</strong></li><li><strong>IT leaders admit they'd like to move, but they just can't handle the complexity</strong></li></ul><p>We're already familiar with lifting workloads out of certain environments in favor of digital sovereignty, but new Akamai research has revealed that half of enterprises are running workloads in suboptimal cloud locations despite knowing that alternatives offer better performance or cost.</p><p>The report notes that traditional cloud architecture is starting to struggle with modern applications, including real-time APIs, AI and more.</p><p>In fact, four in five (82%) enterprises are now missing response time targets for business-critical interactive workloads – but they might be reluctant to change just out of pure simplicity.</p><h2 id="enterprises-don-39-t-want-the-hassle-of-moving-cloud-providers">Enterprises don't want the hassle of moving cloud providers</h2><p>Akamai found that storing and processing data in different geographic locations (41%) and longer-than-ideal physical distances between end users and compute (38%) are to blame for these response time issues. But for many, accepting these trade-offs might be more acceptable than migrating workloads to an entirely new cloud vendor.</p><p>"As real-time interactions, conversational AI, and microservices become standard, running compute far from where data lives creates delays and inefficiencies that organizations can no longer sustain," Akamai Cloud Product Marketing VP Ari Weil explained.</p><p>Four in five (79%) worry doing this could take days, or more, but it's the investments into building applications around the current cloud provider that could be the biggest headache. Two in five (38%) say workloads would need to be re-architected before they can be relocated, with 8% worried they wouldn't even be able to move workloads at all once they've been deployed.</p><p>Half of the enterprises Akamai surveyed admitted they've left workloads in non-ideal geographic locations just because relocating them would be too expensive, complicated or disruptive. But if it were easier, three-quarters of IT leaders say they'd deploy compute resources to more geographic locations to reduce latency, improve performance and expand into new markets.</p><p>Akamai can't exactly urge enterprises to consider alternative cloud vendors and infrastructure given the circumstances, but the report certainly highlights how interconnected and dependent our tech stacks have become.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/half-of-businesses-run-workloads-in-the-wrong-place-mainly-because-they-cant-be-bothered-to-change-their-ways</link>
                                                                            <description>
                            <![CDATA[ Businesses are admitting that current cloud infrastructure is holding them back, but they also can't face the challenges involved with migrating. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">QM3gQ5qmAKSkzd7KJDNXB7</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/ky3eJEUVLiJuSkHTgVNnLf-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 15:17:43 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Cloud Computing]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/ky3eJEUVLiJuSkHTgVNnLf-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Man tapping a cloud icon]]></media:description>                                                            <media:text><![CDATA[Man tapping a cloud icon]]></media:text>
                                <media:title type="plain"><![CDATA[Man tapping a cloud icon]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/ky3eJEUVLiJuSkHTgVNnLf-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Unsuitable cloud infrastructure is adding latency and reducing performance</strong></li><li><strong>Enterprises worry they'd have to re-architecture applications just to move</strong></li><li><strong>IT leaders admit they'd like to move, but they just can't handle the complexity</strong></li></ul><p>We're already familiar with lifting workloads out of certain environments in favor of digital sovereignty, but new Akamai research has revealed that half of enterprises are running workloads in suboptimal cloud locations despite knowing that alternatives offer better performance or cost.</p><p>The report notes that traditional cloud architecture is starting to struggle with modern applications, including real-time APIs, AI and more.</p><p>In fact, four in five (82%) enterprises are now missing response time targets for business-critical interactive workloads – but they might be reluctant to change just out of pure simplicity.</p><h2 id="enterprises-don-39-t-want-the-hassle-of-moving-cloud-providers">Enterprises don't want the hassle of moving cloud providers</h2><p>Akamai found that storing and processing data in different geographic locations (41%) and longer-than-ideal physical distances between end users and compute (38%) are to blame for these response time issues. But for many, accepting these trade-offs might be more acceptable than migrating workloads to an entirely new cloud vendor.</p><p>"As real-time interactions, conversational AI, and microservices become standard, running compute far from where data lives creates delays and inefficiencies that organizations can no longer sustain," Akamai Cloud Product Marketing VP Ari Weil explained.</p><p>Four in five (79%) worry doing this could take days, or more, but it's the investments into building applications around the current cloud provider that could be the biggest headache. Two in five (38%) say workloads would need to be re-architected before they can be relocated, with 8% worried they wouldn't even be able to move workloads at all once they've been deployed.</p><p>Half of the enterprises Akamai surveyed admitted they've left workloads in non-ideal geographic locations just because relocating them would be too expensive, complicated or disruptive. But if it were easier, three-quarters of IT leaders say they'd deploy compute resources to more geographic locations to reduce latency, improve performance and expand into new markets.</p><p>Akamai can't exactly urge enterprises to consider alternative cloud vendors and infrastructure given the circumstances, but the report certainly highlights how interconnected and dependent our tech stacks have become.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Why general-purpose AI falls short in agriculture ]]></title>
                                                                                                <dc:content><![CDATA[ <p>In July, a farmer in Chuzhou, China, turned to an <a href="https://www.techradar.com/pro/best-ai-chatbot-for-business">AI chatbot</a> for help. He needed a plan to control weeds and pests in his fields. The model produced what a Futurism journalist later called “a Long Island iced tea of agrochemicals.” The farmer sprayed this cocktail across 25 acres of crops. By the next day, every single sesame seedling was dead, costing him roughly 150,000 yuan, or about 18,000 euros.</p><p>When Wu asked the chatbot what went wrong, the AI admitted it made mistakes. That mix it proposed was suitable for broadleaf weeds. But sesame is widely classified as a broadleaf plant, so it was never the right option to begin with. Another error was suggesting spraying the entire field instead of using the targeted application. </p><p>As horrible as it is, that is just one example of AI missing the mark. Researchers from the U.S., the U.K., Kenya, Nigeria, and Colombia recently tested whether generative AI could replace human agricultural advisors. Apparently, it can’t. </p><p>According to their paper published in Nature Food, the team first asked chatbots to recommend treatment for the fall armyworm, a pest that destroys corn crops worldwide. In response, OpenAI's GPT-3.5 and GPT-4 models provided highly ambiguous pesticide advice. </p><p>Next, the researchers simulated questions from cassava farmers in Nigeria. ChatGPT recommended specific herbicides but gave the wrong application timing. Following that advice would have killed the crops, threatening food security for millions of people, since Nigeria is Africa's most important cassava producer. </p><h2 id="more-frequent-mistakes">More frequent mistakes</h2><p>What's more disturbing is that these mistakes might become more frequent. With rising labor, land, and equipment costs, many farmers can no longer afford professional advice. On top of that, there is a severe shortage of trained agricultural extensionists. Extension workers rarely reach the world's most isolated fields, such as those in South America and Africa, while farmers in those regions cannot travel hundreds of kilometers to attend seminars to improve their knowledge. </p><p>Instead, they often rely on extension agencies based in capital cities. But those lack both the funding and the staff to update their outdated guides or track new farming trends. So farmers turn to digital assistants. According to a recent McKinsey survey, roughly 17 percent of farmers worldwide already use generative <a href="https://www.techradar.com/best/best-ai-tools">AI</a> in day-to-day decisions. That is more than any other new technology in the sector, including robotics and electric-powered machinery, Bloomberg reported.</p><p>To be fair, AI itself is not the issue. The problem is that many farmers turn to general-purpose AI chatbots such as ChatGPT or Claude for advice. These tools know crop types, chemical classes, standard practices, and textbook planting calendars. But they are not trained specifically for agriculture purposes. For example, they won't ask for necessary data from the user before providing advice. As a result, their answers can sound professional while being useless or even harmful.</p><h2 id="training-ai-for-agronomic-purposes">Training AI for agronomic purposes</h2><p>There is a way to fix it: train an AI model for agronomic purposes. That takes two steps. First, the model learns from real <a href="https://www.techradar.com/pro/best-data-removal-services-of-year">data</a> about a particular field: satellite images, past harvests, current weather and the crop's stage of growth. Second, it learns from working agronomists, who teach it how to respond in different situations. Those lessons also set limits on what the model will say. If it knows what crop is in the ground, it won't recommend a herbicide that is banned for that crop.</p><p>To better understand how this works, imagine a farmer who needs to spray a wheat field to get rid of weeds. A standard AI chatbot might draw on general textbooks and suggest a strong chemical. But because it cannot see the actual field, it does not know two critical things: that the wheat is at a fragile stage of growth where this chemical will ruin the crop, and that tomorrow's high winds will blow the spray onto neighboring land. Following the AI's advice could destroy the harvest.</p><p>By contrast, an AI connected to real data checks satellite <a href="https://www.techradar.com/best/best-ways-to-backup-photos">photos</a> and weather reports first. It automatically spots the high winds, recognizes that the wheat is too fragile for certain chemicals, and picks a safe alternative. It also asks questions when information is unclear or missing. Then it shows the farmer exactly where to spray, using coordinates or a task map that targets only the weed patches. In the end, this saves money and protects future yields.</p><h2 id="collecting-information">Collecting information</h2><p>Having said all that, expecting farmers to gather all this information and enter it into a generic AI tool isn't realistic. It takes time most growers don't have, as well as technical knowledge. Plus, farmers have no time to do inputs regularly. </p><p>This is why I believe a new market will emerge: of specialized agricultural AI assistants with built-in access to dashboards and satellite imagery, allowing seamless movement between data sources and making accurate, field-specific information far easier to access and act on. </p><p>This market is already taking shape. Some assistants can revisit satellite data layers, analyze them visually, and answer questions like “What has changed in the last 48 hours?” in plain <a href="https://www.techradar.com/best/best-language-learning-apps">language</a>. Some products also let farmers send voice messages describing what they have applied. The agent files those details in the right parts of the system and keeps them as context. This makes it easier to keep field records up to date, which helps the model give more useful answers.</p><p>Such tools can help farmers make sense of their own field data, spot meaningful changes, and spend less time piecing together multiple sources. With some luck, a wave of these purpose-built tools will help eliminate costly mistakes in agriculture, so more farmers worldwide can thrive instead of losing their crops, money and spirit.</p><p><em></em><a href="https://www.techradar.com/best/best-business-cloud-storage-service"><em>We've reviewed and ranked the best business cloud storage services</em></a><em>.</em></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/why-general-purpose-ai-falls-short-in-agriculture</link>
                                                                            <description>
                            <![CDATA[ AI can transform farming, but without field-specific data, its advice risks costly mistakes. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">YGgo7BJieSCi3CJxgucrKg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/PAztEScphfxGJfYno5NjrL-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 15:02:00 +0000</pubDate>                                                                                                                                <updated>Tue, 17 Nov 2026 15:07:03 +0000</updated>
                                                                                                                                            <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Katya Kheistver ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/PAztEScphfxGJfYno5NjrL-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A robot standing thoughtfully in front of a giant digital display with code on it]]></media:description>                                                            <media:text><![CDATA[A robot standing thoughtfully in front of a giant digital display with code on it]]></media:text>
                                <media:title type="plain"><![CDATA[A robot standing thoughtfully in front of a giant digital display with code on it]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/PAztEScphfxGJfYno5NjrL-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>In July, a farmer in Chuzhou, China, turned to an <a href="https://www.techradar.com/pro/best-ai-chatbot-for-business">AI chatbot</a> for help. He needed a plan to control weeds and pests in his fields. The model produced what a Futurism journalist later called “a Long Island iced tea of agrochemicals.” The farmer sprayed this cocktail across 25 acres of crops. By the next day, every single sesame seedling was dead, costing him roughly 150,000 yuan, or about 18,000 euros.</p><p>When Wu asked the chatbot what went wrong, the AI admitted it made mistakes. That mix it proposed was suitable for broadleaf weeds. But sesame is widely classified as a broadleaf plant, so it was never the right option to begin with. Another error was suggesting spraying the entire field instead of using the targeted application. </p><p>As horrible as it is, that is just one example of AI missing the mark. Researchers from the U.S., the U.K., Kenya, Nigeria, and Colombia recently tested whether generative AI could replace human agricultural advisors. Apparently, it can’t. </p><p>According to their paper published in Nature Food, the team first asked chatbots to recommend treatment for the fall armyworm, a pest that destroys corn crops worldwide. In response, OpenAI's GPT-3.5 and GPT-4 models provided highly ambiguous pesticide advice. </p><p>Next, the researchers simulated questions from cassava farmers in Nigeria. ChatGPT recommended specific herbicides but gave the wrong application timing. Following that advice would have killed the crops, threatening food security for millions of people, since Nigeria is Africa's most important cassava producer. </p><h2 id="more-frequent-mistakes">More frequent mistakes</h2><p>What's more disturbing is that these mistakes might become more frequent. With rising labor, land, and equipment costs, many farmers can no longer afford professional advice. On top of that, there is a severe shortage of trained agricultural extensionists. Extension workers rarely reach the world's most isolated fields, such as those in South America and Africa, while farmers in those regions cannot travel hundreds of kilometers to attend seminars to improve their knowledge. </p><p>Instead, they often rely on extension agencies based in capital cities. But those lack both the funding and the staff to update their outdated guides or track new farming trends. So farmers turn to digital assistants. According to a recent McKinsey survey, roughly 17 percent of farmers worldwide already use generative <a href="https://www.techradar.com/best/best-ai-tools">AI</a> in day-to-day decisions. That is more than any other new technology in the sector, including robotics and electric-powered machinery, Bloomberg reported.</p><p>To be fair, AI itself is not the issue. The problem is that many farmers turn to general-purpose AI chatbots such as ChatGPT or Claude for advice. These tools know crop types, chemical classes, standard practices, and textbook planting calendars. But they are not trained specifically for agriculture purposes. For example, they won't ask for necessary data from the user before providing advice. As a result, their answers can sound professional while being useless or even harmful.</p><h2 id="training-ai-for-agronomic-purposes">Training AI for agronomic purposes</h2><p>There is a way to fix it: train an AI model for agronomic purposes. That takes two steps. First, the model learns from real <a href="https://www.techradar.com/pro/best-data-removal-services-of-year">data</a> about a particular field: satellite images, past harvests, current weather and the crop's stage of growth. Second, it learns from working agronomists, who teach it how to respond in different situations. Those lessons also set limits on what the model will say. If it knows what crop is in the ground, it won't recommend a herbicide that is banned for that crop.</p><p>To better understand how this works, imagine a farmer who needs to spray a wheat field to get rid of weeds. A standard AI chatbot might draw on general textbooks and suggest a strong chemical. But because it cannot see the actual field, it does not know two critical things: that the wheat is at a fragile stage of growth where this chemical will ruin the crop, and that tomorrow's high winds will blow the spray onto neighboring land. Following the AI's advice could destroy the harvest.</p><p>By contrast, an AI connected to real data checks satellite <a href="https://www.techradar.com/best/best-ways-to-backup-photos">photos</a> and weather reports first. It automatically spots the high winds, recognizes that the wheat is too fragile for certain chemicals, and picks a safe alternative. It also asks questions when information is unclear or missing. Then it shows the farmer exactly where to spray, using coordinates or a task map that targets only the weed patches. In the end, this saves money and protects future yields.</p><h2 id="collecting-information">Collecting information</h2><p>Having said all that, expecting farmers to gather all this information and enter it into a generic AI tool isn't realistic. It takes time most growers don't have, as well as technical knowledge. Plus, farmers have no time to do inputs regularly. </p><p>This is why I believe a new market will emerge: of specialized agricultural AI assistants with built-in access to dashboards and satellite imagery, allowing seamless movement between data sources and making accurate, field-specific information far easier to access and act on. </p><p>This market is already taking shape. Some assistants can revisit satellite data layers, analyze them visually, and answer questions like “What has changed in the last 48 hours?” in plain <a href="https://www.techradar.com/best/best-language-learning-apps">language</a>. Some products also let farmers send voice messages describing what they have applied. The agent files those details in the right parts of the system and keeps them as context. This makes it easier to keep field records up to date, which helps the model give more useful answers.</p><p>Such tools can help farmers make sense of their own field data, spot meaningful changes, and spend less time piecing together multiple sources. With some luck, a wave of these purpose-built tools will help eliminate costly mistakes in agriculture, so more farmers worldwide can thrive instead of losing their crops, money and spirit.</p><p><em></em><a href="https://www.techradar.com/best/best-business-cloud-storage-service"><em>We've reviewed and ranked the best business cloud storage services</em></a><em>.</em></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Nearly half of LinkedIn users would like to start their own business — with younger workers particularly keen ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Gen Z and Millennials are among the most likely to start their own businesses</strong></li><li><strong>More founders are becoming content creators on social media networks</strong></li><li><strong>AI can help run your business, but human authenticity is sought-after by customers</strong></li></ul><p>Almost half (46%) of global respondents in a recent LinkedIn survey said they wanted to start their own business – a figure that grows to two in three (63%) Millennials and three in four (72%) Gen Z.</p><p>But aspirations aside, the social networking platform also revealed that one in two Gen Z even plan to l launch their business within the next three years, demonstrating that action is being taken to fulfill dreams.</p><p>Consequently, the platform has seen a 125% increase in the number of Founder profiles in the US since 2022 – or a 275% increase among US Gen Z respondents.</p><h2 id="linkedin-says-more-young-people-are-starting-their-own-businesses">LinkedIn says more young people are starting their own businesses</h2><p>LinkedIn thanked AI for lowering many barriers to entry, with 53% agreeing there are more opportunities to become an entrepreneur today than just five years ago. An overwhelming majority (92%) of small business owners now use AI, with benefits including faster launches (28%), launching with less capital (27%) and eliminating the need for a technical co-founder (27%).</p><p>But novel technologies aside, LinkedIn also says platforms like itself also play a role – 60% said social media makes it easier to find and reach customers where they are, and 42% noted that digital platforms have helped them to establish reputations.</p><p>Putting the two concepts together, LinkedIn also argues that professional networks have become more valuable as AI adoption has grown – something that four in five (78%) small business owners agreed with. Founders are especially fond of human networks when it comes to seeking peer advice on critical business decisions.</p><h2 id="entrepreneurs-are-increasingly-turning-to-multiple-revenue-streams">Entrepreneurs are increasingly turning to multiple revenue streams</h2><p>The report found that 70% of SMB owners now have multiple revenue streams, which can often include one primary business and other smaller streams of revenue through consultancy or freelance work. Greater control and flexibility (47%) and higher earnings potential (41%) were among the biggest influencers for extra work.</p><p>Video is also continuing to see considerable growth, with 73% of founders now describing themselves as content creators, up from 59% in a study just six months earlier. This could be a consequence of an increasingly digital, AI-driven world, in which consumers are seeking more human connections. Three in four (77%) founders agree their personal brand directly contributes credibility to their business.</p><p>In other words, we're living in a business world where AI assistance and human authenticity not only can, but must, coexist. </p><p>"Our research shows the relationships you create, the reputation you build and the people you can turn to for advice become even more valuable as AI helps support the everyday mechanics of a business," Senior Director for Economic Research Sharat Raghavan explained.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/nearly-half-of-linkedin-users-would-like-to-start-their-own-business-with-younger-workers-particularly-keen</link>
                                                                            <description>
                            <![CDATA[ Younger generations are starting their own businesses for greater control, flexibility and earning potential, LinkedIn says. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">p7ueSomJXDhj596NnU7XQ8</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/tVU8k3Apu3kqrTTGSzyYGd-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 14:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/tVU8k3Apu3kqrTTGSzyYGd-1920-80.jpg">
                                                            <media:credit><![CDATA[Lance Ulanoff / Future]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Apple iPhone Duo REVIEW]]></media:description>                                                            <media:text><![CDATA[Apple iPhone Duo REVIEW]]></media:text>
                                <media:title type="plain"><![CDATA[Apple iPhone Duo REVIEW]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/tVU8k3Apu3kqrTTGSzyYGd-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Gen Z and Millennials are among the most likely to start their own businesses</strong></li><li><strong>More founders are becoming content creators on social media networks</strong></li><li><strong>AI can help run your business, but human authenticity is sought-after by customers</strong></li></ul><p>Almost half (46%) of global respondents in a recent LinkedIn survey said they wanted to start their own business – a figure that grows to two in three (63%) Millennials and three in four (72%) Gen Z.</p><p>But aspirations aside, the social networking platform also revealed that one in two Gen Z even plan to l launch their business within the next three years, demonstrating that action is being taken to fulfill dreams.</p><p>Consequently, the platform has seen a 125% increase in the number of Founder profiles in the US since 2022 – or a 275% increase among US Gen Z respondents.</p><h2 id="linkedin-says-more-young-people-are-starting-their-own-businesses">LinkedIn says more young people are starting their own businesses</h2><p>LinkedIn thanked AI for lowering many barriers to entry, with 53% agreeing there are more opportunities to become an entrepreneur today than just five years ago. An overwhelming majority (92%) of small business owners now use AI, with benefits including faster launches (28%), launching with less capital (27%) and eliminating the need for a technical co-founder (27%).</p><p>But novel technologies aside, LinkedIn also says platforms like itself also play a role – 60% said social media makes it easier to find and reach customers where they are, and 42% noted that digital platforms have helped them to establish reputations.</p><p>Putting the two concepts together, LinkedIn also argues that professional networks have become more valuable as AI adoption has grown – something that four in five (78%) small business owners agreed with. Founders are especially fond of human networks when it comes to seeking peer advice on critical business decisions.</p><h2 id="entrepreneurs-are-increasingly-turning-to-multiple-revenue-streams">Entrepreneurs are increasingly turning to multiple revenue streams</h2><p>The report found that 70% of SMB owners now have multiple revenue streams, which can often include one primary business and other smaller streams of revenue through consultancy or freelance work. Greater control and flexibility (47%) and higher earnings potential (41%) were among the biggest influencers for extra work.</p><p>Video is also continuing to see considerable growth, with 73% of founders now describing themselves as content creators, up from 59% in a study just six months earlier. This could be a consequence of an increasingly digital, AI-driven world, in which consumers are seeking more human connections. Three in four (77%) founders agree their personal brand directly contributes credibility to their business.</p><p>In other words, we're living in a business world where AI assistance and human authenticity not only can, but must, coexist. </p><p>"Our research shows the relationships you create, the reputation you build and the people you can turn to for advice become even more valuable as AI helps support the everyday mechanics of a business," Senior Director for Economic Research Sharat Raghavan explained.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ New report finds nearly all UK data centre operators report electrical disruption as 'critical bottleneck' looms ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>The UK has a 6GW target for AI capacity by 2030, but its data centers aren't moving quickly enough</strong></li><li><strong>The average connection wait time has grown to 14 months, up from 12 months globally</strong></li><li><strong>One in four say they only discovered an electrical problem after a failure or downtime</strong></li></ul><p>Four in five (79%) UK data center operators have blamed electrical or power-system issues for major disruptions, delays or downtime, leaving just 17% confident that they're fully-prepared for future AI workloads, new Fluke research has claimed.</p><p>And these challenges are already impacting operations, with around a third (32%) reporting that power problems have already limited their abilities to deploy high-density GPUs and AI workloads.</p><p>This is especially alarming given that the UK is targeting 6GW of data center capacity by the end of the decade spanning multiple large-scale AI Growth Zones.</p><h2 id="uk-data-centers-are-struggling-with-electricity-supplies">UK data centers are struggling with electricity supplies</h2><p>With electrical infrastructure now seen as one of the biggest bottlenecks, only 41% say they have real-time visibility across their critical electrical systems. A further quarter (26%) say they only discovered electrical problems after a failure or downtime.</p><p>But it's not just the power infrastructure that's problematic – many operators are struggling to hire enough talent, with half (51%) admitting recruiting electrical technicians has become harder. Two in five (38%) say this has directly delayed infrastructure upgrades because they don't have the right people to lead an upgrade team.</p><p>As for new projects or expansions, operators are now seeing an average of 14 months to wait for additional grid connections or power capacity – two months longer than the UK average.</p><p>"As AI drives unprecedented demand for computing capacity, the electrical infrastructure inside data centres is becoming a critical bottleneck," Fluke Corporation President Parker Burke summarized.</p><p>Looking ahead, the fact that the UK is falling behind the global average poses serious questions as to whether it can deliver on its 2030 AI targets. With grid constraints and skills shortages both threatening fast enough expansion, a more harmonious strategy involving the energy sector is clearly needed.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/new-report-finds-nearly-all-uk-data-centre-operators-report-electrical-disruption-as-critical-bottleneck-looms</link>
                                                                            <description>
                            <![CDATA[ The average new or expanded UK data center is waiting 14 months for an electrical supply, compared with 12 months globally. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">LJSUevrVefrhehYXSTN9tC</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/aBt4xDYv6hbiB3JyuDnEuM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 13:15:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/aBt4xDYv6hbiB3JyuDnEuM-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A digital grid criss-crossing the lights of a city below]]></media:description>                                                            <media:text><![CDATA[A digital grid criss-crossing the lights of a city below]]></media:text>
                                <media:title type="plain"><![CDATA[A digital grid criss-crossing the lights of a city below]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/aBt4xDYv6hbiB3JyuDnEuM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>The UK has a 6GW target for AI capacity by 2030, but its data centers aren't moving quickly enough</strong></li><li><strong>The average connection wait time has grown to 14 months, up from 12 months globally</strong></li><li><strong>One in four say they only discovered an electrical problem after a failure or downtime</strong></li></ul><p>Four in five (79%) UK data center operators have blamed electrical or power-system issues for major disruptions, delays or downtime, leaving just 17% confident that they're fully-prepared for future AI workloads, new Fluke research has claimed.</p><p>And these challenges are already impacting operations, with around a third (32%) reporting that power problems have already limited their abilities to deploy high-density GPUs and AI workloads.</p><p>This is especially alarming given that the UK is targeting 6GW of data center capacity by the end of the decade spanning multiple large-scale AI Growth Zones.</p><h2 id="uk-data-centers-are-struggling-with-electricity-supplies">UK data centers are struggling with electricity supplies</h2><p>With electrical infrastructure now seen as one of the biggest bottlenecks, only 41% say they have real-time visibility across their critical electrical systems. A further quarter (26%) say they only discovered electrical problems after a failure or downtime.</p><p>But it's not just the power infrastructure that's problematic – many operators are struggling to hire enough talent, with half (51%) admitting recruiting electrical technicians has become harder. Two in five (38%) say this has directly delayed infrastructure upgrades because they don't have the right people to lead an upgrade team.</p><p>As for new projects or expansions, operators are now seeing an average of 14 months to wait for additional grid connections or power capacity – two months longer than the UK average.</p><p>"As AI drives unprecedented demand for computing capacity, the electrical infrastructure inside data centres is becoming a critical bottleneck," Fluke Corporation President Parker Burke summarized.</p><p>Looking ahead, the fact that the UK is falling behind the global average poses serious questions as to whether it can deliver on its 2030 AI targets. With grid constraints and skills shortages both threatening fast enough expansion, a more harmonious strategy involving the energy sector is clearly needed.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ The world's most popular passwords for 2026 have been revealed — and unsurprisingly, they're all pretty bad ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Weak passwords like “123456”, “admin”, and “12345678” remain among the world’s most common choices</strong></li><li><strong>Researchers warn popular sports stars, games, and internet trends are also frequently used and easily cracked</strong></li><li><strong>Password managers, passkeys, and multi-factor authentication remain the best defenses against account compromise</strong></li></ul><p>We’re just over a month away from 2027 and people still choose ‘12345678’ as their password for some reason.</p><p>This is according to NordPass’ latest (and utterly depressing) <a href="https://nordpass.com/yPNR2weqT6HmxVXbgqJdkG4vbS5tCYvD/" target="_blank" rel="nofollow">report</a> on the most popular passwords people use around the world. </p><p>The company has once again aggregated all of the passwords criminals leak on the dark web, analyzed them, and tells the world what the most common passwords and phrases are. And year after year, the report shows the same five passwords (or password variant, should we say) being at the very top:</p><p>‘12345668’ or a variant (‘1234’, ‘123456’, or similar), ‘password’, and ‘admin’. </p><h2 id="why-are-we-like-this">Why are we like this?</h2><p>“Once again, “123456” is the most commonly used password in the world. That makes it seven out of eight years this password has stood at the top of the leaderboard,” NordPass says in its report. </p><p>“In fact, the top three remain unchanged, with “admin” and “12345678” taking the second and third places respectively. We still have a lot of work to do to spread awareness about password safety.”</p><p>The report also devotes plenty of attention to people using their favorite players and pop culture characters in their passwords. </p><p>“Quality over quantity — although video games and anime made up the largest share of entries in our pop culture passwords list, sports-related passwords were used the most,” the report continues. Lebron James was the name that popped up the most in passwords, followed by Messi, Lampard, and Lewandowski. Curiously enough, no Ronaldo, Travis Kelce, or Tom Brady.</p><p>Savage, Roblox, Skibidi, Covid, and Impostor, were also frequently used words. </p><p>The resilience of simple passwords is astounding, especially given that many online services nowadays require users to create strong passwords - ones that combine uppercase and lowercase letters, numbers, and special characters. </p><p>For years, security researchers have been warning that passwords are the weakest link in the cybersecurity chain and that users should move to more secure solutions, such as biometrics. At the very least, they should activate multi-factor authentication (MFA) wherever possible.</p><h2 id="fandoms-make-terrible-passwords">Fandoms make terrible passwords</h2><p>Today, the risk is even greater, the report found.</p><p>The advent of AI has made cracking passwords that much faster, easier, and more streamlined, and the tool has also significantly sped up the rate at which login credentials are being stolen. </p><p>Not a week goes by without at least one report of millions of passwords being stolen and leaked somewhere. Many are encrypted, but considering how fast AI and quantum computing are progressing, this is more of a roadblock and less of a serious hindrance for cybercriminals. </p><p>Using popular characters doesn’t help, either. “Fandoms make for terrible passwords,” NordPass says. “The one thing that connects these password choices? They all could be cracked in seconds.”</p><p>Unfortunately, it seems that we’ve learned very little about password hygiene over the years, and users still prefer simple, easy-to-remember passwords over the safety of their online accounts and the data they hold. </p><p>To make matters worse, these risks can easily be mitigated, either by moving to passkeys and biometric options (which are offered by most services these days), or by simply installing one of many free-to-use password managers. These tools can generate strong passwords with a single click and can even autofill them, in case convenience is users’ number one concern. </p><p>Really, in 2027, there is no excuse for having such a weak password. Get it together, people!</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/the-worlds-most-popular-passwords-for-2026-have-been-revealed-and-unsurprisingly-theyre-all-pretty-bad</link>
                                                                            <description>
                            <![CDATA[ A variant of '123456' is still the most popular password in the world - eight years in. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">hsrQ8UbD8wF4UQfJafQywg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/WnLTf8xmKy47sCbjkg2qH5-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 12:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/WnLTf8xmKy47sCbjkg2qH5-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A sticky note on a laptop reading &quot;password 12345678&quot;]]></media:description>                                                            <media:text><![CDATA[A sticky note on a laptop reading &quot;password 12345678&quot;]]></media:text>
                                <media:title type="plain"><![CDATA[A sticky note on a laptop reading &quot;password 12345678&quot;]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/WnLTf8xmKy47sCbjkg2qH5-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Weak passwords like “123456”, “admin”, and “12345678” remain among the world’s most common choices</strong></li><li><strong>Researchers warn popular sports stars, games, and internet trends are also frequently used and easily cracked</strong></li><li><strong>Password managers, passkeys, and multi-factor authentication remain the best defenses against account compromise</strong></li></ul><p>We’re just over a month away from 2027 and people still choose ‘12345678’ as their password for some reason.</p><p>This is according to NordPass’ latest (and utterly depressing) <a href="https://nordpass.com/yPNR2weqT6HmxVXbgqJdkG4vbS5tCYvD/" target="_blank" rel="nofollow">report</a> on the most popular passwords people use around the world. </p><p>The company has once again aggregated all of the passwords criminals leak on the dark web, analyzed them, and tells the world what the most common passwords and phrases are. And year after year, the report shows the same five passwords (or password variant, should we say) being at the very top:</p><p>‘12345668’ or a variant (‘1234’, ‘123456’, or similar), ‘password’, and ‘admin’. </p><h2 id="why-are-we-like-this">Why are we like this?</h2><p>“Once again, “123456” is the most commonly used password in the world. That makes it seven out of eight years this password has stood at the top of the leaderboard,” NordPass says in its report. </p><p>“In fact, the top three remain unchanged, with “admin” and “12345678” taking the second and third places respectively. We still have a lot of work to do to spread awareness about password safety.”</p><p>The report also devotes plenty of attention to people using their favorite players and pop culture characters in their passwords. </p><p>“Quality over quantity — although video games and anime made up the largest share of entries in our pop culture passwords list, sports-related passwords were used the most,” the report continues. Lebron James was the name that popped up the most in passwords, followed by Messi, Lampard, and Lewandowski. Curiously enough, no Ronaldo, Travis Kelce, or Tom Brady.</p><p>Savage, Roblox, Skibidi, Covid, and Impostor, were also frequently used words. </p><p>The resilience of simple passwords is astounding, especially given that many online services nowadays require users to create strong passwords - ones that combine uppercase and lowercase letters, numbers, and special characters. </p><p>For years, security researchers have been warning that passwords are the weakest link in the cybersecurity chain and that users should move to more secure solutions, such as biometrics. At the very least, they should activate multi-factor authentication (MFA) wherever possible.</p><h2 id="fandoms-make-terrible-passwords">Fandoms make terrible passwords</h2><p>Today, the risk is even greater, the report found.</p><p>The advent of AI has made cracking passwords that much faster, easier, and more streamlined, and the tool has also significantly sped up the rate at which login credentials are being stolen. </p><p>Not a week goes by without at least one report of millions of passwords being stolen and leaked somewhere. Many are encrypted, but considering how fast AI and quantum computing are progressing, this is more of a roadblock and less of a serious hindrance for cybercriminals. </p><p>Using popular characters doesn’t help, either. “Fandoms make for terrible passwords,” NordPass says. “The one thing that connects these password choices? They all could be cracked in seconds.”</p><p>Unfortunately, it seems that we’ve learned very little about password hygiene over the years, and users still prefer simple, easy-to-remember passwords over the safety of their online accounts and the data they hold. </p><p>To make matters worse, these risks can easily be mitigated, either by moving to passkeys and biometric options (which are offered by most services these days), or by simply installing one of many free-to-use password managers. These tools can generate strong passwords with a single click and can even autofill them, in case convenience is users’ number one concern. </p><p>Really, in 2027, there is no excuse for having such a weak password. Get it together, people!</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ GoDaddy's new AI builder app lets SMBs turn an idea into a website faster than ever ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>GoDaddy's AI website builder already supports 90+ third-party integrations</strong></li><li><strong>You can build and operate an ecommerce site with payment platform support</strong></li><li><strong>Early usage data shows it being a hit among service-based businesses</strong></li></ul><p>GoDaddy has launched Airo AI Builder, its own agentic operating system designed to create advanced websites.</p><p>Positioned as much more than a setup tool, Airo will instead let users both create and continue to refine their sites via a simple conversational interface, with website spanning both simple portfolios as well as web apps, online stores, online booking systems and more.</p><p>From launch, the agentic system already supports integrations with over 90 third-party tools including Slack, Notion and HubSpot, focusing on tasks like accessing databases and managing customer relationships.</p><h2 id="godaddy-launches-airo-ai-builder">GoDaddy launches Airo AI Builder</h2><p>The company recently released research which found service-based businesses accounted for more than one-third of the sites created, including health and body (13%), home and property (12%) and business services (11%) sectors. Food and travel (13%) also accounted for a large number of sites created, all of which are not typically associated with having extensive tech teams or expertise – making effective website building a challenge for many.</p><p>During Airo AI Builder's beta phase in the summer, around one-tenth of the use cases even included rebuild and reimagine projects – businesses looking to update or modernise their sites, who may not previously have had the right skills or resources.</p><p>On the ecommerce front, GoDaddy offers its own Payments platform but it also has a partnership with Stripe to let users collect payments, but more are set to come.</p><p>"For anyone who has found building a website a burden, it opens up the opportunity to take that first step, then keep improving as their customers’ needs and their business evolve," marketing-led growth VP Selina Bieber said, positioning Airo as a more comprehensive platform designed to see users through more than just the initial build.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/godaddys-new-ai-builder-app-lets-smbs-turn-an-idea-into-a-website-faster-than-ever</link>
                                                                            <description>
                            <![CDATA[ Airo AI Builder is GoDaddy's comprehensive AI platform designed for building and editing your ideal website. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">kKQWpLNfNbqj95oGV9rkBK</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/LcFmUSZ8zro2hdwA7MwsmR-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 10:49:59 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Website Building]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/LcFmUSZ8zro2hdwA7MwsmR-1920-80.png">
                                                            <media:credit><![CDATA[GoDaddy]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A website being built on GoDaddy&#039;s Airo site builder.]]></media:description>                                                            <media:text><![CDATA[A website being built on GoDaddy&#039;s Airo site builder.]]></media:text>
                                <media:title type="plain"><![CDATA[A website being built on GoDaddy&#039;s Airo site builder.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/LcFmUSZ8zro2hdwA7MwsmR-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>GoDaddy's AI website builder already supports 90+ third-party integrations</strong></li><li><strong>You can build and operate an ecommerce site with payment platform support</strong></li><li><strong>Early usage data shows it being a hit among service-based businesses</strong></li></ul><p>GoDaddy has launched Airo AI Builder, its own agentic operating system designed to create advanced websites.</p><p>Positioned as much more than a setup tool, Airo will instead let users both create and continue to refine their sites via a simple conversational interface, with website spanning both simple portfolios as well as web apps, online stores, online booking systems and more.</p><p>From launch, the agentic system already supports integrations with over 90 third-party tools including Slack, Notion and HubSpot, focusing on tasks like accessing databases and managing customer relationships.</p><h2 id="godaddy-launches-airo-ai-builder">GoDaddy launches Airo AI Builder</h2><p>The company recently released research which found service-based businesses accounted for more than one-third of the sites created, including health and body (13%), home and property (12%) and business services (11%) sectors. Food and travel (13%) also accounted for a large number of sites created, all of which are not typically associated with having extensive tech teams or expertise – making effective website building a challenge for many.</p><p>During Airo AI Builder's beta phase in the summer, around one-tenth of the use cases even included rebuild and reimagine projects – businesses looking to update or modernise their sites, who may not previously have had the right skills or resources.</p><p>On the ecommerce front, GoDaddy offers its own Payments platform but it also has a partnership with Stripe to let users collect payments, but more are set to come.</p><p>"For anyone who has found building a website a burden, it opens up the opportunity to take that first step, then keep improving as their customers’ needs and their business evolve," marketing-led growth VP Selina Bieber said, positioning Airo as a more comprehensive platform designed to see users through more than just the initial build.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ How AI is poised to reshape software teams ]]></title>
                                                                                                <dc:content><![CDATA[ <p>For decades, businesses have organized teams around software. Since starting Alorica in 1999, I’ve observed firsthand in <a href="https://www.techradar.com/best/cx-tools">customer experience</a> (CX) how separate systems for <a href="https://www.techradar.com/best/the-best-crm-software">CRM</a>, ticketing, quality assurance and knowledge have created corresponding roles and responsibilities across IT and operations.</p><p>People are central to these systems — configuring the workflows, managing permissions, maintaining integrations and overseeing data. And operational teams have evolved around the same structure, creating specialist functions for different parts of the customer journey.</p><p>The legacy CX stack was built for a world where humans did the work and software helped them. That world is over. We often discuss what AI means for individuals and jobs. The broader organizational change, though, could unfold around the teams that have grown up in traditional software.</p><p>When you go agentic, all the structure and complexity of the old system falls away. Suddenly, businesses no longer need to organize people around individual applications in the same way.</p><p>Yet humans stay central. The model shifts to governance. AI will operate and compile insights. People remain responsible for oversight, judgement and ensuring AI delivers the right outcomes. They spend less time managing individual workflows and processes.</p><p>As a result, agentic AI is likely to change more than the technology <a href="https://www.techradar.com/best/best-business-cloud-storage-service">businesses</a> use. It could lead to fewer functional silos, broader responsibilities and teams organized around outcomes rather than individual applications.</p><p>Which sparks a big question for businesses: As the current organizational structures built around software lose relevance, what should they look like in the age of agentic AI?</p><h2 id="the-ai-journey-in-cx-software-so-far">The AI journey in CX software so far</h2><p>We see three approaches to AI in <a href="https://www.techradar.com/best/best-database-software">software</a> on the market. The first is bolt-on, where independent AI products are layered on to existing software. I’ve bought almost every major platform in this market and deployed them across thousands of customers. I know where this model fails because I’ve seen it many times. The AI is kneecapped — it can’t do more than the fragmented legacy systems it’s bolted onto. You end up with a faster version of the same problem.</p><p>The second approach is embedding AI models within a product. There are benefits: For one, the administration of the software is easier. Instead of a human having to perform all of the programming to change workflows, when you embed a large language model that is able to program into the software, it can take over some of this work.</p><p>However, this is still a step away from the third, and currently most advanced, approach. A native agentic AI system.</p><p>There’s a clear difference between, for example, an <a href="https://www.techradar.com/computing/artificial-intelligence/best-llms">LLM</a> doing some programming and a fully agentic system. In the agentic model, AI agents orchestrate, learn, and improve every customer interaction across operations. Think Uber or DoorDash — apps like this appear fully automated from the outside.</p><p>But you’d better believe humans are working to close the gaps the technology can't. Agentic AI doesn't delete that need. Instead, it moves the gaps and changes who closes them.</p><h2 id="intelligence-shifting-from-software-to-ai">Intelligence shifting from software to AI</h2><p>These approaches also demonstrate an acknowledgment by the market: Intelligence is moving away from software and into AI. The value of software has degraded to the point where it’s just a repository – all the intelligence now happens in the AI layer.</p><p>This is why companies have transitioned from the first to the second approach. But as the software still needs to be administered and managed, it’s still not the most efficient way of doing things. Transformation lies in the adoption of a native agentic AI system, and that’s where the world is going. Yet in making this move, organizational structures are going to have to change with it. </p><h2 id="organizing-around-outcomes-over-applications">Organizing around outcomes over applications </h2><p>As today’s companies are organized around their software, getting the full value from agentic AI will require them to rethink how their teams and responsibilities are structured. At core, we’re talking about teaching people how to work differently — governing versus doing.</p><p>All of the components within a software platform are performing functions that humans need to manage. But when agents can manage the configuration of these tools, the need for specialists in each area reduces. ‘Broader thinkers’ are key. These people work across components, observing and governing processes at a wider scale. Companies facilitate this shift by creating new or adapted job roles and responsibilities for existing <a href="https://www.techradar.com/pro/best-employee-management-software-of-year">employees</a>.</p><p>The silo is a major enterprise challenge. A department specialist will focus on their area of responsibility and look at things through their perspective. This setup naturally creates tension. If one person owns this tower and another that tower, conflicts are inevitable and collaboration suffers.</p><p>The move to agentic calms those tensions. AI agents figure out the optimal way to perform a task across departments and applications. With human biases mitigated, the best outcomes have a better chance of surfacing. Instead of a specialist assessing how ticketing can be improved, for example, they’re now monitoring a whole agentic system that is learning how to improve every conversation and interaction across the CX. </p><h2 id="who-wins-in-the-cx-future">Who wins in the CX future</h2><p>The companies that benefit most from agentic AI won’t simply be using the best technology. The most powerful enterprise transformation lies in rethinking how work gets done — in how to organize the organization.</p><p>It's going to be disruptive. Companies who initially protect existing structures as they implement agentic AI will hit conflicts and confusion. In the next five to 10 years, the boldest adopters of fully agentic will gain significant competitive advantages. Those who don’t will lose market share — it’s not an if but a when.</p><p><em></em><a href="https://www.techradar.com/best/best-ai-tools"><em>We've featured the best AI tool.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/how-ai-is-poised-to-reshape-software-teams</link>
                                                                            <description>
                            <![CDATA[ The rise of agentic AI could fundamentally change how businesses structure their teams. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">zxeFo4jJyAxguJzQKPSMBf</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/PRCsQfoXPXi2t4jsGwWr6L-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 10:28:08 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Andy Lee ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/PRCsQfoXPXi2t4jsGwWr6L-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A human shakes a robot&#039;s hand in front of blue concentric circles]]></media:description>                                                            <media:text><![CDATA[A human shakes a robot&#039;s hand in front of blue concentric circles]]></media:text>
                                <media:title type="plain"><![CDATA[A human shakes a robot&#039;s hand in front of blue concentric circles]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/PRCsQfoXPXi2t4jsGwWr6L-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>For decades, businesses have organized teams around software. Since starting Alorica in 1999, I’ve observed firsthand in <a href="https://www.techradar.com/best/cx-tools">customer experience</a> (CX) how separate systems for <a href="https://www.techradar.com/best/the-best-crm-software">CRM</a>, ticketing, quality assurance and knowledge have created corresponding roles and responsibilities across IT and operations.</p><p>People are central to these systems — configuring the workflows, managing permissions, maintaining integrations and overseeing data. And operational teams have evolved around the same structure, creating specialist functions for different parts of the customer journey.</p><p>The legacy CX stack was built for a world where humans did the work and software helped them. That world is over. We often discuss what AI means for individuals and jobs. The broader organizational change, though, could unfold around the teams that have grown up in traditional software.</p><p>When you go agentic, all the structure and complexity of the old system falls away. Suddenly, businesses no longer need to organize people around individual applications in the same way.</p><p>Yet humans stay central. The model shifts to governance. AI will operate and compile insights. People remain responsible for oversight, judgement and ensuring AI delivers the right outcomes. They spend less time managing individual workflows and processes.</p><p>As a result, agentic AI is likely to change more than the technology <a href="https://www.techradar.com/best/best-business-cloud-storage-service">businesses</a> use. It could lead to fewer functional silos, broader responsibilities and teams organized around outcomes rather than individual applications.</p><p>Which sparks a big question for businesses: As the current organizational structures built around software lose relevance, what should they look like in the age of agentic AI?</p><h2 id="the-ai-journey-in-cx-software-so-far">The AI journey in CX software so far</h2><p>We see three approaches to AI in <a href="https://www.techradar.com/best/best-database-software">software</a> on the market. The first is bolt-on, where independent AI products are layered on to existing software. I’ve bought almost every major platform in this market and deployed them across thousands of customers. I know where this model fails because I’ve seen it many times. The AI is kneecapped — it can’t do more than the fragmented legacy systems it’s bolted onto. You end up with a faster version of the same problem.</p><p>The second approach is embedding AI models within a product. There are benefits: For one, the administration of the software is easier. Instead of a human having to perform all of the programming to change workflows, when you embed a large language model that is able to program into the software, it can take over some of this work.</p><p>However, this is still a step away from the third, and currently most advanced, approach. A native agentic AI system.</p><p>There’s a clear difference between, for example, an <a href="https://www.techradar.com/computing/artificial-intelligence/best-llms">LLM</a> doing some programming and a fully agentic system. In the agentic model, AI agents orchestrate, learn, and improve every customer interaction across operations. Think Uber or DoorDash — apps like this appear fully automated from the outside.</p><p>But you’d better believe humans are working to close the gaps the technology can't. Agentic AI doesn't delete that need. Instead, it moves the gaps and changes who closes them.</p><h2 id="intelligence-shifting-from-software-to-ai">Intelligence shifting from software to AI</h2><p>These approaches also demonstrate an acknowledgment by the market: Intelligence is moving away from software and into AI. The value of software has degraded to the point where it’s just a repository – all the intelligence now happens in the AI layer.</p><p>This is why companies have transitioned from the first to the second approach. But as the software still needs to be administered and managed, it’s still not the most efficient way of doing things. Transformation lies in the adoption of a native agentic AI system, and that’s where the world is going. Yet in making this move, organizational structures are going to have to change with it. </p><h2 id="organizing-around-outcomes-over-applications">Organizing around outcomes over applications </h2><p>As today’s companies are organized around their software, getting the full value from agentic AI will require them to rethink how their teams and responsibilities are structured. At core, we’re talking about teaching people how to work differently — governing versus doing.</p><p>All of the components within a software platform are performing functions that humans need to manage. But when agents can manage the configuration of these tools, the need for specialists in each area reduces. ‘Broader thinkers’ are key. These people work across components, observing and governing processes at a wider scale. Companies facilitate this shift by creating new or adapted job roles and responsibilities for existing <a href="https://www.techradar.com/pro/best-employee-management-software-of-year">employees</a>.</p><p>The silo is a major enterprise challenge. A department specialist will focus on their area of responsibility and look at things through their perspective. This setup naturally creates tension. If one person owns this tower and another that tower, conflicts are inevitable and collaboration suffers.</p><p>The move to agentic calms those tensions. AI agents figure out the optimal way to perform a task across departments and applications. With human biases mitigated, the best outcomes have a better chance of surfacing. Instead of a specialist assessing how ticketing can be improved, for example, they’re now monitoring a whole agentic system that is learning how to improve every conversation and interaction across the CX. </p><h2 id="who-wins-in-the-cx-future">Who wins in the CX future</h2><p>The companies that benefit most from agentic AI won’t simply be using the best technology. The most powerful enterprise transformation lies in rethinking how work gets done — in how to organize the organization.</p><p>It's going to be disruptive. Companies who initially protect existing structures as they implement agentic AI will hit conflicts and confusion. In the next five to 10 years, the boldest adopters of fully agentic will gain significant competitive advantages. Those who don’t will lose market share — it’s not an if but a when.</p><p><em></em><a href="https://www.techradar.com/best/best-ai-tools"><em>We've featured the best AI tool.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Does the end of vSphere 8 support mean the end of your VMware estate? ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Having supported VMware <a href="https://www.techradar.com/best/best-customer-feedback-tools">customers</a> through both vendor and independent support roles for over a decade, I’ve seen many organizations plan and execute major upgrades. The software change itself is only one part of the work. In a mature VMware estate, an upgrade reaches into applications, integrations, infrastructure dependencies, operational processes, and the teams responsible for keeping critical services available.</p><p>Even a well-understood upgrade can become a substantial operational undertaking. Before a production environment changes, <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> teams need to understand how the wider estate will be affected, what needs to move in what order, where incompatibilities may appear, and how much risk the business is prepared to absorb. In large environments, that work can take months. A migration to another virtualization platform can take considerably longer.</p><p>That long lead time is exactly why the vSphere 8 support deadline matters now. General support ends on 11th October 2027, and for a large enterprise the remaining window may be shorter than the time needed to assess the estate, build a <a href="https://www.techradar.com/best/best-business-cloud-storage-service">business</a> case, test dependencies, and complete a major upgrade or migration safely.</p><p>The deadline creates two pressures at once: organizations need a plan for how vSphere 8 will be supported beyond October 2027, while also deciding whether the longer-term strategy is to upgrade, migrate or continue running parts of the existing estate for longer. For some customers, particularly those running hyperconverged infrastructure, an upgrade may also require a <a href="https://www.techradar.com/news/best-processors">hardware</a> refresh, adding significantly to the cost.</p><p>In practice, the commercial route to continued vendor support can bring those decisions together. Customers may move onto the newer subscription or upgrade path because that is what allows them to remain supported, even when they are not yet ready to make the corresponding technical change across the estate. </p><p>We are already seeing this play out. Some organizations have purchased VMware Cloud Foundation subscriptions but continue to run workloads on vSphere 8 rather than upgrading to vSphere 9. From a commercial perspective, they have moved onto Broadcom's subscription model.</p><p>Technically, however, the estate remains where it was. That apparent contradiction says something important about the decision customers are facing. Purchasing a subscription does not remove the operational work involved in deploying a newer platform, nor does it automatically establish the business case for doing so. In a mature VMware estate, those considerations can take much longer to work through.    </p><p>As 2027 approaches, VMware customers need a credible plan for supporting vSphere 8, alongside a longer-term view of where those workloads should ultimately run. The two decisions influence one another, but forcing them onto the same timetable can push organizations towards change before the business case. Technical readiness or migration plan is sufficiently developed.</p><h2 id="why-the-architecture-decision-deserves-its-own-timetable">Why the architecture decision deserves its own timetable</h2><p>There is a tendency to think of an upgrade as a progression from one version of a product to the next. In enterprise infrastructure, the consequences are rarely so contained.</p><p>A mature VMware estate carries years of operational knowledge around it. Applications have been tested against the platform; backup and recovery processes have developed around it, and infrastructure teams understand how changes in one part of the environment affect another. None of that means the estate should be preserved indefinitely. But it does mean that replacing it carries a cost that extends beyond the new <a href="https://www.techradar.com/best/best-small-business-software">software</a> or platform itself.</p><p>I saw this repeatedly during my time at VMware. Upgrades were one of the biggest drivers of support cases because changing the virtualization layer could expose compatibility problems or dependencies elsewhere in the stack.</p><p>A move may still be entirely justified. The current environment may be becoming harder to secure or maintain. Another platform may offer better value for money, or the organization may have infrastructure requirements that the existing estate can no longer meet.</p><p>For some CIOs, that assessment also includes vendor constraints and lock-in, and whether <a href="https://www.techradar.com/best/the-best-open-source-crm-of-year">open source</a> alternatives or complementary technologies could better support their longer-term virtualization strategy.</p><p>What matters is that those reasons come from the needs of the business and the condition of the environment. An approaching support deadline tells an organization that its current support arrangements need attention. It does not provide the business case for a wider infrastructure transformation.</p><h2 id="supporting-vsphere-8-beyond-the-vendor-deadline">Supporting vSphere 8 beyond the vendor deadline</h2><p>For customers that expect to still be running vSphere 8 after October 2027, the choice is not simply between upgrading and continuing without support. Independent third-party support offers another option: supporting the existing environment while giving the organization time to assess and implement a longer-term platform decision.</p><p>That can be valuable in several different circumstances. A customer that has already decided to leave VMware may need years to migrate a large estate without creating unnecessary operational risk. Another may still be assessing newer VMware options against competing platforms. Some organizations may find that different groups of workloads need to move at different points rather than as part of one wholesale <a href="https://www.techradar.com/best/best-data-migration-tools">migration</a>.</p><p>Third-party support does not make the VMware lifecycle irrelevant, and moving beyond vendor support requires a different approach to managing risk. Without the vendor patch stream, organizations need to understand vulnerabilities in the context of their own environment: how VMware is configured, where it is exposed, and what controls already sit around it.</p><p>That means taking a defense-in-depth approach, using measures such as configuration hardening, segmentation, access controls, and compensating controls where appropriate. The aim is not to suggest that patches no longer matter but to avoid treating them as the only way to manage security risk once an environment moves beyond the vendor support lifecycle.</p><p>What third-party support fundamentally changes is the amount of control the customer has over the timetable.</p><h2 id="using-the-support-deadline-to-make-a-better-decision">Using the support deadline to make a better decision</h2><p>The October 2027 deadline should already be prompting VMware customers to examine their estates in detail. They need to understand how long existing workloads are likely to remain on vSphere 8, what maintaining those environments will require and where there is a genuine case for change.</p><p>For some organizations, that assessment will support an upgrade or migration before general support ends. Others will conclude that the right destination is clear but cannot responsibly be reached within the remaining support window. There will also be environments where the current platform continues to meet the needs of the business and changing it is not yet a priority.</p><p>Each of those positions still requires a support strategy, but the longer-term architecture decision may need to follow a different timetable.</p><p>An alternative support model allows those timelines to be managed separately. The end of vendor support still needs to be planned for, but it does not have to determine when a wider infrastructure change takes place. That decision can instead be based on the technical requirements of the estate, the operational risk involved in changing it and the economics of the available options.</p><p>The support deadline requires a plan. It does not have to dictate the future of the VMware estate.</p><p><em></em><a href="https://www.techradar.com/best/best-cloud-computing-services"><em>We've featured the best cloud computing provider</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/does-the-end-of-vsphere-8-support-mean-the-end-of-your-vmware-estate</link>
                                                                            <description>
                            <![CDATA[ A mature VMware estate carries years of operational knowledge around it. Replacing it carries a cost that extends beyond the new software or platform itself. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">DLWh83FSmkNJxfVetK5uD9</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/wZAaq2s2qH4tHBJTEBNZXM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 09:56:40 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Shane O’Rourke ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/wZAaq2s2qH4tHBJTEBNZXM-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[An abstract pattern of blue lines and orange-yellow dots on a dark blue background, to represent a digital environment]]></media:description>                                                            <media:text><![CDATA[An abstract pattern of blue lines and orange-yellow dots on a dark blue background, to represent a digital environment]]></media:text>
                                <media:title type="plain"><![CDATA[An abstract pattern of blue lines and orange-yellow dots on a dark blue background, to represent a digital environment]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/wZAaq2s2qH4tHBJTEBNZXM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Having supported VMware <a href="https://www.techradar.com/best/best-customer-feedback-tools">customers</a> through both vendor and independent support roles for over a decade, I’ve seen many organizations plan and execute major upgrades. The software change itself is only one part of the work. In a mature VMware estate, an upgrade reaches into applications, integrations, infrastructure dependencies, operational processes, and the teams responsible for keeping critical services available.</p><p>Even a well-understood upgrade can become a substantial operational undertaking. Before a production environment changes, <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> teams need to understand how the wider estate will be affected, what needs to move in what order, where incompatibilities may appear, and how much risk the business is prepared to absorb. In large environments, that work can take months. A migration to another virtualization platform can take considerably longer.</p><p>That long lead time is exactly why the vSphere 8 support deadline matters now. General support ends on 11th October 2027, and for a large enterprise the remaining window may be shorter than the time needed to assess the estate, build a <a href="https://www.techradar.com/best/best-business-cloud-storage-service">business</a> case, test dependencies, and complete a major upgrade or migration safely.</p><p>The deadline creates two pressures at once: organizations need a plan for how vSphere 8 will be supported beyond October 2027, while also deciding whether the longer-term strategy is to upgrade, migrate or continue running parts of the existing estate for longer. For some customers, particularly those running hyperconverged infrastructure, an upgrade may also require a <a href="https://www.techradar.com/news/best-processors">hardware</a> refresh, adding significantly to the cost.</p><p>In practice, the commercial route to continued vendor support can bring those decisions together. Customers may move onto the newer subscription or upgrade path because that is what allows them to remain supported, even when they are not yet ready to make the corresponding technical change across the estate. </p><p>We are already seeing this play out. Some organizations have purchased VMware Cloud Foundation subscriptions but continue to run workloads on vSphere 8 rather than upgrading to vSphere 9. From a commercial perspective, they have moved onto Broadcom's subscription model.</p><p>Technically, however, the estate remains where it was. That apparent contradiction says something important about the decision customers are facing. Purchasing a subscription does not remove the operational work involved in deploying a newer platform, nor does it automatically establish the business case for doing so. In a mature VMware estate, those considerations can take much longer to work through.    </p><p>As 2027 approaches, VMware customers need a credible plan for supporting vSphere 8, alongside a longer-term view of where those workloads should ultimately run. The two decisions influence one another, but forcing them onto the same timetable can push organizations towards change before the business case. Technical readiness or migration plan is sufficiently developed.</p><h2 id="why-the-architecture-decision-deserves-its-own-timetable">Why the architecture decision deserves its own timetable</h2><p>There is a tendency to think of an upgrade as a progression from one version of a product to the next. In enterprise infrastructure, the consequences are rarely so contained.</p><p>A mature VMware estate carries years of operational knowledge around it. Applications have been tested against the platform; backup and recovery processes have developed around it, and infrastructure teams understand how changes in one part of the environment affect another. None of that means the estate should be preserved indefinitely. But it does mean that replacing it carries a cost that extends beyond the new <a href="https://www.techradar.com/best/best-small-business-software">software</a> or platform itself.</p><p>I saw this repeatedly during my time at VMware. Upgrades were one of the biggest drivers of support cases because changing the virtualization layer could expose compatibility problems or dependencies elsewhere in the stack.</p><p>A move may still be entirely justified. The current environment may be becoming harder to secure or maintain. Another platform may offer better value for money, or the organization may have infrastructure requirements that the existing estate can no longer meet.</p><p>For some CIOs, that assessment also includes vendor constraints and lock-in, and whether <a href="https://www.techradar.com/best/the-best-open-source-crm-of-year">open source</a> alternatives or complementary technologies could better support their longer-term virtualization strategy.</p><p>What matters is that those reasons come from the needs of the business and the condition of the environment. An approaching support deadline tells an organization that its current support arrangements need attention. It does not provide the business case for a wider infrastructure transformation.</p><h2 id="supporting-vsphere-8-beyond-the-vendor-deadline">Supporting vSphere 8 beyond the vendor deadline</h2><p>For customers that expect to still be running vSphere 8 after October 2027, the choice is not simply between upgrading and continuing without support. Independent third-party support offers another option: supporting the existing environment while giving the organization time to assess and implement a longer-term platform decision.</p><p>That can be valuable in several different circumstances. A customer that has already decided to leave VMware may need years to migrate a large estate without creating unnecessary operational risk. Another may still be assessing newer VMware options against competing platforms. Some organizations may find that different groups of workloads need to move at different points rather than as part of one wholesale <a href="https://www.techradar.com/best/best-data-migration-tools">migration</a>.</p><p>Third-party support does not make the VMware lifecycle irrelevant, and moving beyond vendor support requires a different approach to managing risk. Without the vendor patch stream, organizations need to understand vulnerabilities in the context of their own environment: how VMware is configured, where it is exposed, and what controls already sit around it.</p><p>That means taking a defense-in-depth approach, using measures such as configuration hardening, segmentation, access controls, and compensating controls where appropriate. The aim is not to suggest that patches no longer matter but to avoid treating them as the only way to manage security risk once an environment moves beyond the vendor support lifecycle.</p><p>What third-party support fundamentally changes is the amount of control the customer has over the timetable.</p><h2 id="using-the-support-deadline-to-make-a-better-decision">Using the support deadline to make a better decision</h2><p>The October 2027 deadline should already be prompting VMware customers to examine their estates in detail. They need to understand how long existing workloads are likely to remain on vSphere 8, what maintaining those environments will require and where there is a genuine case for change.</p><p>For some organizations, that assessment will support an upgrade or migration before general support ends. Others will conclude that the right destination is clear but cannot responsibly be reached within the remaining support window. There will also be environments where the current platform continues to meet the needs of the business and changing it is not yet a priority.</p><p>Each of those positions still requires a support strategy, but the longer-term architecture decision may need to follow a different timetable.</p><p>An alternative support model allows those timelines to be managed separately. The end of vendor support still needs to be planned for, but it does not have to determine when a wider infrastructure change takes place. That decision can instead be based on the technical requirements of the estate, the operational risk involved in changing it and the economics of the available options.</p><p>The support deadline requires a plan. It does not have to dictate the future of the VMware estate.</p><p><em></em><a href="https://www.techradar.com/best/best-cloud-computing-services"><em>We've featured the best cloud computing provider</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Over half of businesses admit they're facing significantly higher AI costs than they expected ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Financial costs of AI are now a cause of concern for over 50% of executives</strong></li><li><strong>While token cost is a concern, related expenditure such as training and governance is also having an impact</strong></li><li><strong>Few organizations are able to link AI use to related business outcomes, leaving them struggling to justify increasing expenditure</strong></li></ul><p>AI expenditure is busting budgets, even with a decrease in token costs for the most popular AI models, new Capgemini research has claimed.</p><p>The research shows over half (53%) of executives are concerned about AI costs, with expenditure on tokens, training, governance, and supplemental spending on human oversight and repeated work all contributing to the uncertainty.</p><p>Meanwhile, 59% of execs have reported that their expected AI related expenditure was surpassed, with 20% of organizations breaching their AI budget target by over 30%. Consequently, according to the report, more businesses are finding that CFOs are increasingly involved in AI spending decisions.</p><h2 id="ai-yield">AI yield</h2><p>500 CXO-level senior executives working at large organizations were polled during September 2026, with the <a href="https://www.capgemini.com/insights/research-library/ai-economics/" target="_blank">report</a> highlighting the growing financial burden of AI. Responses to the growing cost of AI have included not just tighter budgets, but diversifying which AI models are used – this is known as a multi-model strategy. </p><p>Volker Darius, CEO of Capgemini Invent, explains: “Cost discipline alone will not create a profitable AI estate, just as value creation without cost control will not be sustainable.”</p><p>“Organizations need to measure cost, quality and business value together, use the most appropriate model for each task, and account for the human oversight required to deliver outcomes safely. By continuously measuring AI Yield, organizations can make more informed capital-allocation decisions and scale AI with confidence.”</p><p>AI Yield is described as a metric designed to verify and record the value of an outcome that occurred thanks to AI. Smarter measuring of this factor can support requests for increases AI token spend.</p><h2 id="business-outcomes">Business outcomes</h2><p>Multi-model strategies are expected to become more widely employed over the next 12 months, with three in five of the polled execs expecting to adopt. These will typically be selected on cost and capability primarily, as well as factors such as latency and risk.</p><p>Currently, 60% or organizations measure total AI expenditure against budget, but this is expected to drop to 44% over the next 6-12 months. Instead, says Capgemini Research Institute, metrics that measure successful business outcomes from AI use will be prioritized.</p><p>Darius adds: “As AI moves deeper into business workflows, organizations need to stop treating its cost as a technology or procurement issue and start managing AI as a portfolio of economic outcomes.”</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/over-half-of-businesses-admit-theyre-facing-significantly-higher-ai-costs-than-they-expected</link>
                                                                            <description>
                            <![CDATA[ Organizations are spending more on AI despite a drop in token costs, raising questions on how AI expenditure can be kept under control. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">feJ25fVjTgF6cKmjFAKgzg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/dUfubfF2M2AdcnKK69Zy2Y-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 09:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                                                                                    <dc:creator><![CDATA[ Christian Cawley ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/zBDYnjPnB2XPvhKbYX9Kuc-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Christian Cawley has extensive experience as a writer and editor in consumer electronics, IT and entertainment media. He has contributed to TechRadar since 2017 and has been published in Computer Weekly, Linux Format, ComputerActive, and other publications. &lt;/p&gt;&lt;p&gt;Beyond TechRadar, he heads up the team at smart home website Matter Alpha, and writes about retro gaming at Gaming Retro. &lt;/p&gt;&lt;p&gt;Formerly the editor responsible for Linux, Security, Programming, and DIY at MakeUseOf, Christian previously worked as a desktop and software support specialist in the public and private sectors.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/dUfubfF2M2AdcnKK69Zy2Y-1920-80.jpg">
                                                            <media:credit><![CDATA[Xavier Lorenzo / Getty]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Close-up view of man calculating expenses and managing the budget while using a calculator and laptop at home]]></media:description>                                                            <media:text><![CDATA[Close-up view of man calculating expenses and managing the budget while using a calculator and laptop at home]]></media:text>
                                <media:title type="plain"><![CDATA[Close-up view of man calculating expenses and managing the budget while using a calculator and laptop at home]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/dUfubfF2M2AdcnKK69Zy2Y-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Financial costs of AI are now a cause of concern for over 50% of executives</strong></li><li><strong>While token cost is a concern, related expenditure such as training and governance is also having an impact</strong></li><li><strong>Few organizations are able to link AI use to related business outcomes, leaving them struggling to justify increasing expenditure</strong></li></ul><p>AI expenditure is busting budgets, even with a decrease in token costs for the most popular AI models, new Capgemini research has claimed.</p><p>The research shows over half (53%) of executives are concerned about AI costs, with expenditure on tokens, training, governance, and supplemental spending on human oversight and repeated work all contributing to the uncertainty.</p><p>Meanwhile, 59% of execs have reported that their expected AI related expenditure was surpassed, with 20% of organizations breaching their AI budget target by over 30%. Consequently, according to the report, more businesses are finding that CFOs are increasingly involved in AI spending decisions.</p><h2 id="ai-yield">AI yield</h2><p>500 CXO-level senior executives working at large organizations were polled during September 2026, with the <a href="https://www.capgemini.com/insights/research-library/ai-economics/" target="_blank">report</a> highlighting the growing financial burden of AI. Responses to the growing cost of AI have included not just tighter budgets, but diversifying which AI models are used – this is known as a multi-model strategy. </p><p>Volker Darius, CEO of Capgemini Invent, explains: “Cost discipline alone will not create a profitable AI estate, just as value creation without cost control will not be sustainable.”</p><p>“Organizations need to measure cost, quality and business value together, use the most appropriate model for each task, and account for the human oversight required to deliver outcomes safely. By continuously measuring AI Yield, organizations can make more informed capital-allocation decisions and scale AI with confidence.”</p><p>AI Yield is described as a metric designed to verify and record the value of an outcome that occurred thanks to AI. Smarter measuring of this factor can support requests for increases AI token spend.</p><h2 id="business-outcomes">Business outcomes</h2><p>Multi-model strategies are expected to become more widely employed over the next 12 months, with three in five of the polled execs expecting to adopt. These will typically be selected on cost and capability primarily, as well as factors such as latency and risk.</p><p>Currently, 60% or organizations measure total AI expenditure against budget, but this is expected to drop to 44% over the next 6-12 months. Instead, says Capgemini Research Institute, metrics that measure successful business outcomes from AI use will be prioritized.</p><p>Darius adds: “As AI moves deeper into business workflows, organizations need to stop treating its cost as a technology or procurement issue and start managing AI as a portfolio of economic outcomes.”</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ US marijuana farms are becoming AI data centers — they already have the water and power needed, and with weed revenues falling, it seems an ideal switch ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Falling cannabis revenues are pushing growers to reconsider cultivation properties</strong></li><li><strong>Former marijuana farms have infrastructure that could support smaller AI data centers</strong></li><li><strong>AI demand is giving struggling cannabis properties an unexpected commercial purpose</strong></li></ul><p>The expansion of artificial intelligence infrastructure is creating another possible use for facilities originally constructed to support large-scale indoor marijuana production, new reports have claimed.</p><p>At the same time, declining marijuana prices are forcing some American cannabis growers to reconsider the future of their cultivation properties as operating costs continue rising.</p><p>Several former and current operators are therefore examining whether existing cultivation buildings could accommodate small <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a> without requiring entirely new construction.</p><h2 id="existing-infrastructure-makes-cannabis-properties-attractive-to-data-center-operators">Existing infrastructure makes cannabis properties attractive to data center operators</h2><p>Cannabis cultivation requires substantial electricity and water supplies, particularly for indoor operations involving artificial lighting, ventilation, cooling, and humidity regulation.</p><p>These requirements have left some facilities with electrical connections, industrial cooling equipment, and water infrastructure that could reduce the initial requirements for establishing computing operations.</p><p>In Kalamazoo, Michigan, a 32,850-square-foot cultivation property already has electrical capacity of 1.5 megawatts, with considerable room for expansion.</p><p>Its electricity supply could reach 2.5 megawatts within one year, while longer-term improvements could eventually increase available capacity to 15 megawatts.</p><p>The property also has fiber internet connectivity and an advanced heating, ventilation, and air conditioning system suitable for maintaining controlled indoor temperatures.</p><p>These characteristics could make former cultivation facilities attractive to companies seeking smaller computing locations instead of constructing new industrial complexes.</p><p>“For some reason, there's a lot more cannabis grown in Arizona than Arizonans use. And so, a lot of folks are getting out of that business, like myself, at least on the cultivation side. But these facilities can potentially be repurposed to be mini data center nodes,” said Moe Asnani, former owner of two D2 Dispensaries in Tucson, Arizona.</p><h2 id="conversion-still-carries-substantial-obstacles">Conversion still carries substantial obstacles</h2><p>The proposed conversion, sometimes called adaptive reuse, could also help address opposition against constructing large computing complexes in residential communities.</p><p>Tucson City Council member Paul Cunningham considers existing industrial properties a potentially more acceptable alternative to developing entirely new facilities.</p><p>“Data centers are dirty words, and everybody knows that. And no one wants to build a brand-new data center anywhere,” Cunningham said.</p><p>However, converting cultivation buildings does not automatically make them suitable for commercial computing operations.</p><p>Operators must consider electricity prices, fiber connectivity, and taxation before selecting properties.</p><p>Computing facilities can require two to three times the electrical capacity available at many cultivation properties, making infrastructure improvements particularly important.</p><p>Such upgrades could require as much as five years, potentially delaying projects and weakening the financial attraction of purchasing otherwise inexpensive agricultural properties.</p><p>Smaller installations must also contend with restrictive zoning regulations, construction restrictions, neighborhood opposition, and potentially expensive electricity infrastructure improvements.</p><p>Their operating economics could become particularly challenging as AI processors advance, requiring frequent hardware replacements to maintain competitive computing performance.</p><p>Large operators also benefit from purchasing equipment in substantial quantities, whereas smaller installations could face higher individual hardware costs during subsequent upgrades.</p><p>Nevertheless, resistance against enormous computing developments could encourage operators to investigate existing industrial properties, provided their infrastructure satisfies commercial requirements.</p><p>Via <a href="https://www.tomshardware.com/tech-industry/data-centers/cannabis-farmers-consider-joining-ai-bandwagon-due-to-falling-weed-revenue-existing-sites-could-house-mini-data-center-nodes-but-still-face-serious-challenges-during-conversion" target="_blank">Tom's Hardware</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/us-marijuana-farms-are-becoming-ai-data-centers-they-already-have-the-water-and-power-needed-and-with-weed-revenues-falling-it-seems-an-ideal-switch</link>
                                                                            <description>
                            <![CDATA[ Declining marijuana revenues are prompting growers to convert cultivation facilities into smaller data centers, although infrastructure and regulatory obstacles remain. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">pb7iQuQGUNJ3szv6L4Q3QP</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/9EsRkKapSffbykRzXRqD96-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 01:10:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/9EsRkKapSffbykRzXRqD96-1920-80.png">
                                                            <media:credit><![CDATA[The New York Times]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[US indoor marijuana farms may soon be repurposed for data centers]]></media:description>                                                            <media:text><![CDATA[US indoor marijuana farms may soon be repurposed for data centers]]></media:text>
                                <media:title type="plain"><![CDATA[US indoor marijuana farms may soon be repurposed for data centers]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/9EsRkKapSffbykRzXRqD96-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Falling cannabis revenues are pushing growers to reconsider cultivation properties</strong></li><li><strong>Former marijuana farms have infrastructure that could support smaller AI data centers</strong></li><li><strong>AI demand is giving struggling cannabis properties an unexpected commercial purpose</strong></li></ul><p>The expansion of artificial intelligence infrastructure is creating another possible use for facilities originally constructed to support large-scale indoor marijuana production, new reports have claimed.</p><p>At the same time, declining marijuana prices are forcing some American cannabis growers to reconsider the future of their cultivation properties as operating costs continue rising.</p><p>Several former and current operators are therefore examining whether existing cultivation buildings could accommodate small <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a> without requiring entirely new construction.</p><h2 id="existing-infrastructure-makes-cannabis-properties-attractive-to-data-center-operators">Existing infrastructure makes cannabis properties attractive to data center operators</h2><p>Cannabis cultivation requires substantial electricity and water supplies, particularly for indoor operations involving artificial lighting, ventilation, cooling, and humidity regulation.</p><p>These requirements have left some facilities with electrical connections, industrial cooling equipment, and water infrastructure that could reduce the initial requirements for establishing computing operations.</p><p>In Kalamazoo, Michigan, a 32,850-square-foot cultivation property already has electrical capacity of 1.5 megawatts, with considerable room for expansion.</p><p>Its electricity supply could reach 2.5 megawatts within one year, while longer-term improvements could eventually increase available capacity to 15 megawatts.</p><p>The property also has fiber internet connectivity and an advanced heating, ventilation, and air conditioning system suitable for maintaining controlled indoor temperatures.</p><p>These characteristics could make former cultivation facilities attractive to companies seeking smaller computing locations instead of constructing new industrial complexes.</p><p>“For some reason, there's a lot more cannabis grown in Arizona than Arizonans use. And so, a lot of folks are getting out of that business, like myself, at least on the cultivation side. But these facilities can potentially be repurposed to be mini data center nodes,” said Moe Asnani, former owner of two D2 Dispensaries in Tucson, Arizona.</p><h2 id="conversion-still-carries-substantial-obstacles">Conversion still carries substantial obstacles</h2><p>The proposed conversion, sometimes called adaptive reuse, could also help address opposition against constructing large computing complexes in residential communities.</p><p>Tucson City Council member Paul Cunningham considers existing industrial properties a potentially more acceptable alternative to developing entirely new facilities.</p><p>“Data centers are dirty words, and everybody knows that. And no one wants to build a brand-new data center anywhere,” Cunningham said.</p><p>However, converting cultivation buildings does not automatically make them suitable for commercial computing operations.</p><p>Operators must consider electricity prices, fiber connectivity, and taxation before selecting properties.</p><p>Computing facilities can require two to three times the electrical capacity available at many cultivation properties, making infrastructure improvements particularly important.</p><p>Such upgrades could require as much as five years, potentially delaying projects and weakening the financial attraction of purchasing otherwise inexpensive agricultural properties.</p><p>Smaller installations must also contend with restrictive zoning regulations, construction restrictions, neighborhood opposition, and potentially expensive electricity infrastructure improvements.</p><p>Their operating economics could become particularly challenging as AI processors advance, requiring frequent hardware replacements to maintain competitive computing performance.</p><p>Large operators also benefit from purchasing equipment in substantial quantities, whereas smaller installations could face higher individual hardware costs during subsequent upgrades.</p><p>Nevertheless, resistance against enormous computing developments could encourage operators to investigate existing industrial properties, provided their infrastructure satisfies commercial requirements.</p><p>Via <a href="https://www.tomshardware.com/tech-industry/data-centers/cannabis-farmers-consider-joining-ai-bandwagon-due-to-falling-weed-revenue-existing-sites-could-house-mini-data-center-nodes-but-still-face-serious-challenges-during-conversion" target="_blank">Tom's Hardware</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Russian hackers target Romanian servers to steal NATO, EU emails ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Cereal malware secretly searched Romanian government emails for NATO and EU connections</strong></li><li><strong>ESET discovered previously undocumented software linked to Russia's Turla hacking group</strong></li><li><strong>Hackers exploited Microsoft Exchange systems to conceal their email collection activities</strong></li></ul><p>Cyberespionage hackers allegedly linked to Russia infiltrated the Romanian government email servers, seeking correspondence involving NATO and European Union institutions.</p><p>The operation relied on previously unseen malicious software ESET researchers named Cereal, according to the cybersecurity company's published investigation.</p><p>Rather than sending stolen messages immediately to an outside server, the intruders exploited the victim's own <a href="https://www.techradar.com/news/best-email-provider">email</a><strong> </strong>environment to conceal their activity.</p><h2 id="a-covert-toolset-searched-institutional-correspondence">A covert toolset searched institutional correspondence</h2><p>ESET said the intrusion began on December 23, 2025, after investigators had identified the same software in Kazakhstan the previous month.</p><p>The malicious package contained several .NET components capable of receiving instructions, executing code in memory, and examining an infected organization's internal structure.</p><p>One module, called CerealBox, searched messages for addresses associated with NATO and European Union bodies, then applied filters to identify selected correspondence.</p><p>Emails containing Microsoft Word documents or PDF files were among those selected, after which the software packaged and encrypted copies for collection.</p><p>This approach allowed the operators to obtain specific material without indiscriminately copying every message stored within the compromised organization's mail environment.</p><p>ESET described the software as unusual because its methods relied on detailed knowledge of Exchange systems and unconventional .NET execution techniques.</p><p>The researchers said the malicious code also used legitimate server functions, reducing the need for communications with infrastructure controlled directly by the attackers.</p><p>That arrangement could make detection more difficult because security teams might see activity occurring through systems their organizations already depend upon.</p><p>For institutions managing sensitive diplomatic correspondence, <a href="https://www.techradar.com/best/best-secure-email-providers">secure email</a> protections must therefore include scrutiny of internal server activity, not simply suspicious external messages.</p><h2 id="eset-links-the-activity-to-russia-39-s-turla-group">ESET links the activity to Russia's Turla group</h2><p>ESET attributed the operation to Turla with high confidence, pointing to overlapping tools and similarities with malicious programs previously associated with the group.</p><p>Turla is widely assessed by cybersecurity researchers as connected to Russia's Federal Security Service, which handles internal security and intelligence operations.</p><p>The group has a history of pursuing government organizations, including prominent Ukrainian institutions, and has operated across Eastern Europe and Central Asia.</p><p>According to ESET, around 70% of the countries previously affected by Turla were former Soviet republics or former communist states.</p><p>Romania's location on NATO's eastern edge makes the discovery particularly significant, given the country's exposure to security tensions.</p><p>Romanian authorities have previously faced digital attacks, including a 2025 campaign against defense manufacturers supplying equipment or support to Ukraine.</p><p>Shortly after the war in Ukraine started, denial-of-service incidents disrupted websites belonging to Romanian public bodies and financial institutions.</p><p>The latest compromise adds another example of espionage activity directed at institutions in a country supporting Ukraine and belonging to NATO.</p><p>ESET identified the <a href="https://www.techradar.com/best/best-malware-removal">malware,</a> but the company did not disclose how many emails the hackers managed to steal.</p><p>Via <a href="https://www.defensenews.com/global/europe/2026/11/11/romanian-servers-targeted-by-russian-hackers-to-scoop-up-emails-about-nato-eu/" target="_blank" rel="nofollow">Defense News</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/russian-hackers-target-romanian-servers-to-steal-nato-eu-emails</link>
                                                                            <description>
                            <![CDATA[ Turla-linked operation exploited Microsoft Exchange infrastructure to identify, encrypt, and collect selected NATO- and EU-linked emails without disclosing the total number stolen. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">xKnB5pkcb2oa7GQmYtawUX</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/xNVXLpov5puz2Yq2LcBuoj-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Tue, 17 Nov 2026 00:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/xNVXLpov5puz2Yq2LcBuoj-1920-80.jpg">
                                                            <media:credit><![CDATA[Image by Muhammad Ribkhan from Pixabay]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[email]]></media:description>                                                            <media:text><![CDATA[email]]></media:text>
                                <media:title type="plain"><![CDATA[email]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/xNVXLpov5puz2Yq2LcBuoj-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Cereal malware secretly searched Romanian government emails for NATO and EU connections</strong></li><li><strong>ESET discovered previously undocumented software linked to Russia's Turla hacking group</strong></li><li><strong>Hackers exploited Microsoft Exchange systems to conceal their email collection activities</strong></li></ul><p>Cyberespionage hackers allegedly linked to Russia infiltrated the Romanian government email servers, seeking correspondence involving NATO and European Union institutions.</p><p>The operation relied on previously unseen malicious software ESET researchers named Cereal, according to the cybersecurity company's published investigation.</p><p>Rather than sending stolen messages immediately to an outside server, the intruders exploited the victim's own <a href="https://www.techradar.com/news/best-email-provider">email</a><strong> </strong>environment to conceal their activity.</p><h2 id="a-covert-toolset-searched-institutional-correspondence">A covert toolset searched institutional correspondence</h2><p>ESET said the intrusion began on December 23, 2025, after investigators had identified the same software in Kazakhstan the previous month.</p><p>The malicious package contained several .NET components capable of receiving instructions, executing code in memory, and examining an infected organization's internal structure.</p><p>One module, called CerealBox, searched messages for addresses associated with NATO and European Union bodies, then applied filters to identify selected correspondence.</p><p>Emails containing Microsoft Word documents or PDF files were among those selected, after which the software packaged and encrypted copies for collection.</p><p>This approach allowed the operators to obtain specific material without indiscriminately copying every message stored within the compromised organization's mail environment.</p><p>ESET described the software as unusual because its methods relied on detailed knowledge of Exchange systems and unconventional .NET execution techniques.</p><p>The researchers said the malicious code also used legitimate server functions, reducing the need for communications with infrastructure controlled directly by the attackers.</p><p>That arrangement could make detection more difficult because security teams might see activity occurring through systems their organizations already depend upon.</p><p>For institutions managing sensitive diplomatic correspondence, <a href="https://www.techradar.com/best/best-secure-email-providers">secure email</a> protections must therefore include scrutiny of internal server activity, not simply suspicious external messages.</p><h2 id="eset-links-the-activity-to-russia-39-s-turla-group">ESET links the activity to Russia's Turla group</h2><p>ESET attributed the operation to Turla with high confidence, pointing to overlapping tools and similarities with malicious programs previously associated with the group.</p><p>Turla is widely assessed by cybersecurity researchers as connected to Russia's Federal Security Service, which handles internal security and intelligence operations.</p><p>The group has a history of pursuing government organizations, including prominent Ukrainian institutions, and has operated across Eastern Europe and Central Asia.</p><p>According to ESET, around 70% of the countries previously affected by Turla were former Soviet republics or former communist states.</p><p>Romania's location on NATO's eastern edge makes the discovery particularly significant, given the country's exposure to security tensions.</p><p>Romanian authorities have previously faced digital attacks, including a 2025 campaign against defense manufacturers supplying equipment or support to Ukraine.</p><p>Shortly after the war in Ukraine started, denial-of-service incidents disrupted websites belonging to Romanian public bodies and financial institutions.</p><p>The latest compromise adds another example of espionage activity directed at institutions in a country supporting Ukraine and belonging to NATO.</p><p>ESET identified the <a href="https://www.techradar.com/best/best-malware-removal">malware,</a> but the company did not disclose how many emails the hackers managed to steal.</p><p>Via <a href="https://www.defensenews.com/global/europe/2026/11/11/romanian-servers-targeted-by-russian-hackers-to-scoop-up-emails-about-nato-eu/" target="_blank" rel="nofollow">Defense News</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ A new AWS emulator has been built entirely by AI agents to bring its cloud straight to your PC ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>A new local emulator for AWS has been released by Amazon</strong></li><li><strong>Built by AI, 100% of the code was written by agents</strong></li><li><strong>CC0 licensed and available on GitHub, the project is constantly updated and available now</strong></li></ul><p>A new AWS emulator has been launched, with its code entirely written by AI agents. </p><p>The official release was uploaded to GitHub, but none of the code has been reviewed by the (human) AWS team. </p><p>The project, which began as a “weekend experiment” was inspired by Anthropic’s demonstration that AI agents could build a C compiler autonomously.</p><h2 id="weekend-project">Weekend project</h2><p>The AWS team, using a clearly defined specification, instructed its Kiro autonomous agent with a simple prompt – and were surprised by the results.</p><p>Although not an emulator in the strictest sense, the AWS Emulator provides a useful local platform to help developers become confident that their code will run effectively when uploaded to the real AWS. </p><p>The prompt used to develop the AWS Emulator began life one weekend as a three-step instruction essentially to test Sr. Principal Engineer Matthew Goldberg’s initial hypothesis. </p><p>The agent would perform write tests the AWS Command Line Interface, and iterate on the tests until they can pass on the real AWS. When that was achieved, the agent would then assemble an implementation that the tests would pass with.</p><p>Not only was a passable version built, it took just 15 minutes. By refining the prompt, Goldberg had a version that his team could use by Monday morning.</p><p>“Turning that into a robust product took more than a weekend,” <a href="https://www.aboutamazon.com/news/aws/how-autonomous-ai-agents-built-aws-emulator" target="_blank">wrote</a> Goldberg. “There’s a natural temptation these days to see results like those and believe that AI can build anything. Maybe it will eventually. But building something that you feel confident putting in your customers’ hands still requires serious engineering effort. In the case of AWS Emulator, it required building guardrails that ensure the agents color within our lines.”</p><h2 id="not-quite-an-emulator">Not quite an emulator</h2><p>Despite the billing, the AWS Emulator isn’t strictly speaking an emulator. Instead, it acts as a local application that offers API endpoints replicated from the AWS cloud. As such, not all features of the AWS cloud are available using the tool. However, it is wholly capable of enabling developers to test software without full deployment.</p><p>The <a href="https://github.com/aws/aws-emulator" target="_blank">GitHub repository</a> notes explain “You use it to test your software without calling AWS, on your own machine, on a development server, or in continuous integration.”</p><p>Installing the emulator is straightforward. Mac and PC users need to run it within Docker, while Linux users can access the emulator as an RPM or Debian package, according to their distro. Whichever platform is used, AWS Emulator offers a good chunk of local API access to help developers iron out any problems and ensure that their code will run on a production AWS environment.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/a-new-aws-emulator-has-been-built-entirely-by-ai-agents-to-bring-its-cloud-straight-to-your-pc</link>
                                                                            <description>
                            <![CDATA[ AWS Emulator enables local hosting of AWS via a GitHub download – and the project is 100% coded by AI agents. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">s5qeWoRrAxqSkCqNvViZJm</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/ABdYnex3UvLSEgFZqq9eSZ-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 23:35:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                    <category><![CDATA[Cloud Computing]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Christian Cawley ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/zBDYnjPnB2XPvhKbYX9Kuc-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Christian Cawley has extensive experience as a writer and editor in consumer electronics, IT and entertainment media. He has contributed to TechRadar since 2017 and has been published in Computer Weekly, Linux Format, ComputerActive, and other publications. &lt;/p&gt;&lt;p&gt;Beyond TechRadar, he heads up the team at smart home website Matter Alpha, and writes about retro gaming at Gaming Retro. &lt;/p&gt;&lt;p&gt;Formerly the editor responsible for Linux, Security, Programming, and DIY at MakeUseOf, Christian previously worked as a desktop and software support specialist in the public and private sectors.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/ABdYnex3UvLSEgFZqq9eSZ-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock / nikkimeel]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A hand holding a phone with the AWS logo in front of the AWS logo on an orange background]]></media:description>                                                            <media:text><![CDATA[A hand holding a phone with the AWS logo in front of the AWS logo on an orange background]]></media:text>
                                <media:title type="plain"><![CDATA[A hand holding a phone with the AWS logo in front of the AWS logo on an orange background]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/ABdYnex3UvLSEgFZqq9eSZ-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>A new local emulator for AWS has been released by Amazon</strong></li><li><strong>Built by AI, 100% of the code was written by agents</strong></li><li><strong>CC0 licensed and available on GitHub, the project is constantly updated and available now</strong></li></ul><p>A new AWS emulator has been launched, with its code entirely written by AI agents. </p><p>The official release was uploaded to GitHub, but none of the code has been reviewed by the (human) AWS team. </p><p>The project, which began as a “weekend experiment” was inspired by Anthropic’s demonstration that AI agents could build a C compiler autonomously.</p><h2 id="weekend-project">Weekend project</h2><p>The AWS team, using a clearly defined specification, instructed its Kiro autonomous agent with a simple prompt – and were surprised by the results.</p><p>Although not an emulator in the strictest sense, the AWS Emulator provides a useful local platform to help developers become confident that their code will run effectively when uploaded to the real AWS. </p><p>The prompt used to develop the AWS Emulator began life one weekend as a three-step instruction essentially to test Sr. Principal Engineer Matthew Goldberg’s initial hypothesis. </p><p>The agent would perform write tests the AWS Command Line Interface, and iterate on the tests until they can pass on the real AWS. When that was achieved, the agent would then assemble an implementation that the tests would pass with.</p><p>Not only was a passable version built, it took just 15 minutes. By refining the prompt, Goldberg had a version that his team could use by Monday morning.</p><p>“Turning that into a robust product took more than a weekend,” <a href="https://www.aboutamazon.com/news/aws/how-autonomous-ai-agents-built-aws-emulator" target="_blank">wrote</a> Goldberg. “There’s a natural temptation these days to see results like those and believe that AI can build anything. Maybe it will eventually. But building something that you feel confident putting in your customers’ hands still requires serious engineering effort. In the case of AWS Emulator, it required building guardrails that ensure the agents color within our lines.”</p><h2 id="not-quite-an-emulator">Not quite an emulator</h2><p>Despite the billing, the AWS Emulator isn’t strictly speaking an emulator. Instead, it acts as a local application that offers API endpoints replicated from the AWS cloud. As such, not all features of the AWS cloud are available using the tool. However, it is wholly capable of enabling developers to test software without full deployment.</p><p>The <a href="https://github.com/aws/aws-emulator" target="_blank">GitHub repository</a> notes explain “You use it to test your software without calling AWS, on your own machine, on a development server, or in continuous integration.”</p><p>Installing the emulator is straightforward. Mac and PC users need to run it within Docker, while Linux users can access the emulator as an RPM or Debian package, according to their distro. Whichever platform is used, AWS Emulator offers a good chunk of local API access to help developers iron out any problems and ensure that their code will run on a production AWS environment.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Quote of the day by computer scientist Judea Pearl: "Deep learning has instead given us machines with truly impressive abilities but no intelligence" ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Deep learning has driven several key breakthroughs in AI, powering capabilities like image recognition, multilingual translation, and even real-time conversational partners. </p><p>But despite these impressive demonstrations, real intelligence still eludes our most advanced systems, which continue to excel in what's essentially an advanced form of curve fitting. </p><h2 id="fake-it-until-you-make-it">Fake it until you make it</h2><p>The esteemed computer scientist Judea Pearl was reflecting on the original vision of intelligent machines with the AI systems that we have today in the first chapter of his book.</p><div  class="fancy-box"><div class="fancy_box-title">Quote of the day</div><div class="fancy_box_body"><p class="fancy-box__body-text">This article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. <a data-analytics-id="inline-link" href="https://www.techradar.com/tag/qotd">Read the full series here</a>.</p></div></div><p>Released just a year after <a href="https://research.google/pubs/attention-is-all-you-need/" target="_blank" rel="nofollow">Google scientists' seminal paper</a> outlining the transformer architecture that underpins modern large language models (LLMs), 'The Book of Why' explores how causality changed science. Pearl was a leading AI pioneer and helped devise the probabilistic approach at the heart of deep learning – going on to receive the 2011 Turing Award for his contributions to the field.  </p><p>His thoughts, on the whole, are that AI systems aren't anywhere close to true intelligence despite what comes across as outstanding mimicry, and that they're effectively limited due to a fundamental lack of understanding of cause and effect – instead relying on statistical patterns from which they generate outputs. </p><h2 id="dawn-of-superintelligence">Dawn of superintelligence</h2><p>Despite Pearl insisting that existing AI systems aren't anywhere close to reaching artificial general intelligence (AGI) or similarly advanced milestones, many AI execs have recently insisted that we've either reached that point or that we're close to doing so. </p><p>OpenAI president and co-founder Greg Brockman and ex-Nvidia CEO Jensen Huang both claimed we were at AGI with the launch of recent AI models. Then, we were on the cusp of impressive models like the GPT-6 Astra family and Google's Gemini 4 models. But they still adhered to the same frameworks and architectures that have powered similar LLMs.  </p><p>US President Donald Trump also declared that the term 'artificial intelligence' was redundant because we had now reached '<a href="https://www.techradar.com/ai-platforms-assistants/trump-wants-to-rename-artificial-intelligence-super-intelligence-but-his-real-message-is-that-he-doesnt-want-to-regulate-it-and-thats-what-worries-me" target="_blank">super intelligence</a>' – but ASI is already defined as a concept in science, and there's no basis for this milestone being reached.  </p><div style="min-height: 250px;">                                <div class="kwizly-quiz kwizly-OdvAJe"></div>                            </div>                            <script src="https://kwizly.com/embed/OdvAJe.js" async=""></script> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/quote-of-the-day-by-computer-scientist-judea-pearl-deep-learning-has-instead-given-us-machines-with-truly-impressive-abilities-but-no-intelligence-capturing-the-central-fault-line-in-modern-ai</link>
                                                                            <description>
                            <![CDATA[ Although LLMs and similar systems have impressed us, they haven't yet demonstrated real humanlike smarts ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">Q5qrrRqCXCQ4PTC54if4kC</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/aFci9crB3A9mATCjJ3CWBf-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 23:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Keumars Afifi-Sabet ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/baEeYWYTHEpvddufVqymoA-320-70.jpeg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Keumars Afifi-Sabet is a freelance contributor for Tech Radar and Technology Editor for Live Science. He has written for a variety of publications including ITPro, The Week Digital and ComputerActive. He has worked as a technology journalist for more than five years, having previously held the role of features editor with ITPro. In his previous role, he oversaw the commissioning and publishing of long form in areas including AI, cyber security, cloud computing and digital transformation.&lt;/p&gt;&lt;p&gt;An NCTJ-qualified journalist who specialises in technology, his path into journalism began at university. He immersed himself in student media while studying for a degree in biomedical sciences at Queen Mary, University of London. After graduating, Keumars wrote for a variety of local and national publications as a freelancer, including The Independent, The Observer, and Metro. While studying for his NCTJ certification, his work was commended in the category of ‘Top Scoop’ in the 2017 NCTJ awards. He’s also registered as a foundational chartered manager with the Chartered Management Institute (CMI), having qualified as a Level 3 Team leader with distinction in 2023.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/aFci9crB3A9mATCjJ3CWBf-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images/Amanda Edwards]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Judea Pearl posing for a photograph]]></media:description>                                                            <media:text><![CDATA[Judea Pearl posing for a photograph]]></media:text>
                                <media:title type="plain"><![CDATA[Judea Pearl posing for a photograph]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/aFci9crB3A9mATCjJ3CWBf-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Deep learning has driven several key breakthroughs in AI, powering capabilities like image recognition, multilingual translation, and even real-time conversational partners. </p><p>But despite these impressive demonstrations, real intelligence still eludes our most advanced systems, which continue to excel in what's essentially an advanced form of curve fitting. </p><h2 id="fake-it-until-you-make-it">Fake it until you make it</h2><p>The esteemed computer scientist Judea Pearl was reflecting on the original vision of intelligent machines with the AI systems that we have today in the first chapter of his book.</p><div  class="fancy-box"><div class="fancy_box-title">Quote of the day</div><div class="fancy_box_body"><p class="fancy-box__body-text">This article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. <a data-analytics-id="inline-link" href="https://www.techradar.com/tag/qotd">Read the full series here</a>.</p></div></div><p>Released just a year after <a href="https://research.google/pubs/attention-is-all-you-need/" target="_blank" rel="nofollow">Google scientists' seminal paper</a> outlining the transformer architecture that underpins modern large language models (LLMs), 'The Book of Why' explores how causality changed science. Pearl was a leading AI pioneer and helped devise the probabilistic approach at the heart of deep learning – going on to receive the 2011 Turing Award for his contributions to the field.  </p><p>His thoughts, on the whole, are that AI systems aren't anywhere close to true intelligence despite what comes across as outstanding mimicry, and that they're effectively limited due to a fundamental lack of understanding of cause and effect – instead relying on statistical patterns from which they generate outputs. </p><h2 id="dawn-of-superintelligence">Dawn of superintelligence</h2><p>Despite Pearl insisting that existing AI systems aren't anywhere close to reaching artificial general intelligence (AGI) or similarly advanced milestones, many AI execs have recently insisted that we've either reached that point or that we're close to doing so. </p><p>OpenAI president and co-founder Greg Brockman and ex-Nvidia CEO Jensen Huang both claimed we were at AGI with the launch of recent AI models. Then, we were on the cusp of impressive models like the GPT-6 Astra family and Google's Gemini 4 models. But they still adhered to the same frameworks and architectures that have powered similar LLMs.  </p><p>US President Donald Trump also declared that the term 'artificial intelligence' was redundant because we had now reached '<a href="https://www.techradar.com/ai-platforms-assistants/trump-wants-to-rename-artificial-intelligence-super-intelligence-but-his-real-message-is-that-he-doesnt-want-to-regulate-it-and-thats-what-worries-me" target="_blank">super intelligence</a>' – but ASI is already defined as a concept in science, and there's no basis for this milestone being reached.  </p><div style="min-height: 250px;">                                <div class="kwizly-quiz kwizly-OdvAJe"></div>                            </div>                            <script src="https://kwizly.com/embed/OdvAJe.js" async=""></script>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Tiny Corp's refurbished MI300X tinybox packs 768GB of HBM into a desktop box, and the price has already jumped by $29,000 ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Tiny Corp's quad-GPU MI300X tinybox red sold its 25-unit, $60,000 first batch within three days</strong></li><li><strong>The cards are refurbished, come without warranties, are capped at 500W, and are linked only over PCIe 5.0</strong></li><li><strong>A 50-unit second batch is now on preorder at $89,000, about 48% more, shipping Q1 or Q2 2027</strong></li></ul><p>Tiny Corp has squeezed four of AMD's Instinct MI300X data center accelerators into a box that runs off ordinary wall power.</p><p>The first batch sold out within three days, and the second costs nearly 50% more than the former.</p><p>The <a href="https://x.com/__tinygrad__/status/2118202233145929884" target="_blank" rel="nofollow">tinybox red 4xMI300X went up for preorder on November 5 2026</a> at $60,000, with 25 units in the first batch; but a few days later, <a href="https://x.com/AIQuanting/status/2119310297626530010" target="_blank" rel="nofollow">a breakdown posted by the AI Quanting account on X</a> noted that the store page already showed it as out of stock.</p><h2 id="data-center-silicon-coming-to-a-pricier-desktop-near-you">Data center silicon coming to a pricier desktop near you?</h2><p>Tiny Corp is now taking preorders for a second batch of 50 units at $89,000, about 48% above the launch price, with delivery penciled in for Q1 or Q2 2027, though Tiny Corp stops just short of committing to either as a firm timeline.</p><p>On paper, the numbers are hard to argue with: 768GB of GPU memory, 21.2TB/s of aggregate memory bandwidth and 10.44 PFLOPS of FP8 compute. Those figures are exactly four times <a href="https://www.amd.com/en/products/accelerators/instinct/mi300/mi300x.html" target="_blank" rel="nofollow">AMD's own per-card specifications</a>: 192GB of HBM3, 5.3 TB/s, and 2.61 PFLOPS of dense FP8.</p><p>The listing is also more conservative to boot: Tiny Corp quotes the dense figure rather than the 5.22 PFLOPS AMD lists with structured sparsity.</p><p>The host itself is a capable 32-core, water-cooled AMD EPYC Genoa platform with 192GB of DDR5 RAM, a 1TB boot drive, four 1TB drives in RAID 0, two 10GbE ports, and an OCP 3.0 slot. Two standard 1,700W ATX power supplies feed the 54V the GPUs need through a converter, so the box runs on one 220V circuit or two 110V circuits. Tiny Corp claims it stays under 50dB, which would make it far quieter than the rack servers MI300X cards normally live in.</p><p>Interestingly, these are not the MI300X modules AMD sells to hyperscalers. Tiny Corp refurbishes the cards itself, and they carry no warranty or support from AMD. Each sits on a carrier board Tiny Corp designed, with its own BMC microcontroller, rather than on a standard OAM module.</p><p>There are no XGMI links between the GPUs; each gets a dedicated PCIe 5.0 x16 link to the host instead. AMD's reference design gives every MI300X eight Infinity Fabric links at 128GB/s apiece, so in this box all GPU-to-GPU traffic travels a much narrower road. How that impacts performance is not yet known, with Tiny Corp offering no benchmarks.</p><p>Power is the other quiet compromise. The cards are capped at 500W via a 600W connector, compared to AMD's 750W board power rating. The 10.44 PFLOPS headline is AMD's peak at full power, so with each card running at two-thirds of that budget, sustained compute should theoretically land below the chipmaker's published figure.s </p><p>Tiny Corp hasn't published sustained numbers yet; however, memory bandwidth should suffer less from a power cap, and for single-user inference, that is the figure that matters most.</p><h2 id="a-great-deal-despite-tradeoffs">A great deal despite tradeoffs?</h2><p>The company is unusually candid about the tradeoffs. It warns that the software may be rough at first, but insists "none of the issues should be anything a good agent can't handle." </p><p>On X, it called the configuration plainly unsupported by AMD, while <a href="https://x.com/__tinygrad__/status/2118202833942286381" target="_blank" rel="nofollow">a companion post offered "huge thanks to AMD for supporting us"</a>. The two statements could run in parallel here if you read it as AMD having backed the company, not this particular build.</p><p>The listing also promises the cards will have no uncorrectable HBM errors. That guarantee matters more than it might seem: <a href="https://newsletter.semianalysis.com/p/can-amd-break-the-cuda-moat-amd-advancing" target="_blank" rel="nofollow">SemiAnalysis has reported</a> that Microsoft walked away from AMD after the MI300X, partly due to unreliable Samsung HBM from 2023. Tiny Corp hasn't said where its refurbished cards came from, so buyers are invoking a lot of trust here.</p><p>The MI300X box caps a remarkable reversal. In early 2024, Tiny Corp's first <a href="https://www.techradar.com/pro/another-startup-that-will-cause-gaming-gpu-prices-to-spike-ai-firm-claims-radeon-rx-7900-xtx-gpus-are-better-value-than-nvidias-h100-nearly-six-hundred-backers-believe-that-is-the-case" target="_blank">$15,000 tinybox with six Radeon RX 7900 XTX cards drew 583 preorders</a>. Within weeks, <a href="https://www.tomshardware.com/pc-components/gpus/amd-tinybox-project-put-on-hold-due-to-gpu-instability-in-ai-workloads-firm-publicly-considering-using-intel-gpus" target="_blank" rel="nofollow">the AMD version was put on hold</a> as the company publicly complained that the driver was still very unstable, before it <a href="https://www.phoronix.com/news/Tinybox-Radeon-Again-UMR" target="_blank" rel="nofollow">reversed course days later</a>. A year later, Hotz's <a href="https://geohot.github.io/blog/jekyll/update/2025/03/08/AMD-YOLO.html" target="_blank" rel="nofollow">"AMD YOLO" blog post</a> announced that AMD was shipping the two MI300X systems Tiny Corp had requested.</p><p>Tiny Corp also <a href="https://www.techradar.com/pro/tinygrad-refreshes-tinybox-with-an-old-epyc-cpu-running-ddr4-ram-to-feed-four-amd-9070xt-slaps-a-usd11-000-price-tag-on-a-new-computer-for-deep-learning" target="_blank">relaunched the tinybox in October around four Radeon RX 9070 XT cards for $11,000</a>, and weeks later <a href="https://www.techradar.com/pro/you-cant-currently-order-a-tinybox-with-nvidia-gpus-and-tiny-corps-new-192gb-six-gpu-amd-build-just-got-usd3-000-pricier" target="_blank">its configurator turned AMD-only after a commit titled "rip nvidia"</a>.</p><p>Despite the price hike, the secondhand market makes the box look cheap: Single used and refurbished MI300X OAM modules listed on eBay, as <a href="https://gpupoet.com/gpu/shop/amd-instinct-mi300x" target="_blank" rel="nofollow">tracked by GPU Poet</a>, were asking between about $27,550 and $40,300 on November 16, before anyone supplies a baseboard, host, or power. Even at $89,000, the tinybox works out to about $22,250 per GPU with the rest of the computer included, though asking prices on eBay are not sale prices.</p><p>Not everyone can buy one, either. The listing says the system is export-controlled and will only ship from San Diego to Tier 1 countries, meaning the US and its close allies. Orders are capped at two per customer, and buyers have 7 days to wire funds before an order is canceled.</p><p>A 48% price rise in under a week says plenty about demand for large pools of HBM outside the cloud. But batch two buyers are paying in 2026 for 2027 delivery of warranty-free, refurbished hardware in a configuration AMD doesn't support, and the tokens-per-second figure that would settle the value question has yet to be published officially by the company.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/tiny-corps-refurbished-mi300x-tinybox-packs-768gb-of-hbm-into-a-desktop-box-and-the-price-has-already-jumped-by-usd29-000</link>
                                                                            <description>
                            <![CDATA[ Tiny Corp's $60,000 four-MI300X tinybox sold out in three days; batch two now costs an eye-watering $89,000. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">WiJb7siaWJsVAcfPryrwhg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/n9QrZ8mixWzcJKcRWrBeY4-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 22:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                <author><![CDATA[ Rahimnoorali11@gmail.com (Rahim Amir) ]]></author>                    <dc:creator><![CDATA[ Rahim Amir ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/9xKZFBamtEZKSChRvywbPB-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rahim Amir is a UAE-based tech writer who enjoys building PCs as much as he enjoys writing about them. He has been professionally writing about PC hardware since 2023, focusing on buyer’s guides, hardware reviews, and sponsored content and features related to tech.&lt;br&gt;&lt;br&gt;Having built hundreds of gaming PCs and being an avid gamer in his spare time, Rahim tends to have stronger opinions about hardware than most. This is particularly on display when he gets his way with powerful, but minimalistic RGB builds even as Small Form Factor (SFF) PCs come a close second.&lt;br&gt;&lt;br&gt;In addition to his contributions to TechRadar, Rahim’s work has also been featured on Game Rant and financial news websites.&lt;br&gt;&lt;br&gt;When he’s not working, you can find him playing DotA with friends or schmoozing to take the world over in Civilization. Alternatively, you can find him binging through the entirety of the Lord of The Rings universe with extended editions in play where applicable.&lt;br&gt;&lt;br&gt;You can currently catch Rahim grinding Path of Exile 2, complaining about his (extremely low) unique loot drop rate, or actively participating in one of the numerous (and heated) debates centered around Tolkien&#039;s universe on multiple forums daily.&lt;br&gt;&lt;br&gt;If you have a PC build or a Satisfactory playthrough in progress, he is likely to have some advice to send your way, especially regarding verticality being key for the latter. For the former, Rahim enjoys all aspects of the process including researching the components he will eventually use, benchmarking the latest and greatest hardware he can get his hands on, and somewhat surprisingly, cable management once he gets his latest build to POST.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/n9QrZ8mixWzcJKcRWrBeY4-1920-80.jpg">
                                                            <media:credit><![CDATA[tiny]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A render of the tinybox red]]></media:description>                                                            <media:text><![CDATA[A render of the tinybox red]]></media:text>
                                <media:title type="plain"><![CDATA[A render of the tinybox red]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/n9QrZ8mixWzcJKcRWrBeY4-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Tiny Corp's quad-GPU MI300X tinybox red sold its 25-unit, $60,000 first batch within three days</strong></li><li><strong>The cards are refurbished, come without warranties, are capped at 500W, and are linked only over PCIe 5.0</strong></li><li><strong>A 50-unit second batch is now on preorder at $89,000, about 48% more, shipping Q1 or Q2 2027</strong></li></ul><p>Tiny Corp has squeezed four of AMD's Instinct MI300X data center accelerators into a box that runs off ordinary wall power.</p><p>The first batch sold out within three days, and the second costs nearly 50% more than the former.</p><p>The <a href="https://x.com/__tinygrad__/status/2118202233145929884" target="_blank" rel="nofollow">tinybox red 4xMI300X went up for preorder on November 5 2026</a> at $60,000, with 25 units in the first batch; but a few days later, <a href="https://x.com/AIQuanting/status/2119310297626530010" target="_blank" rel="nofollow">a breakdown posted by the AI Quanting account on X</a> noted that the store page already showed it as out of stock.</p><h2 id="data-center-silicon-coming-to-a-pricier-desktop-near-you">Data center silicon coming to a pricier desktop near you?</h2><p>Tiny Corp is now taking preorders for a second batch of 50 units at $89,000, about 48% above the launch price, with delivery penciled in for Q1 or Q2 2027, though Tiny Corp stops just short of committing to either as a firm timeline.</p><p>On paper, the numbers are hard to argue with: 768GB of GPU memory, 21.2TB/s of aggregate memory bandwidth and 10.44 PFLOPS of FP8 compute. Those figures are exactly four times <a href="https://www.amd.com/en/products/accelerators/instinct/mi300/mi300x.html" target="_blank" rel="nofollow">AMD's own per-card specifications</a>: 192GB of HBM3, 5.3 TB/s, and 2.61 PFLOPS of dense FP8.</p><p>The listing is also more conservative to boot: Tiny Corp quotes the dense figure rather than the 5.22 PFLOPS AMD lists with structured sparsity.</p><p>The host itself is a capable 32-core, water-cooled AMD EPYC Genoa platform with 192GB of DDR5 RAM, a 1TB boot drive, four 1TB drives in RAID 0, two 10GbE ports, and an OCP 3.0 slot. Two standard 1,700W ATX power supplies feed the 54V the GPUs need through a converter, so the box runs on one 220V circuit or two 110V circuits. Tiny Corp claims it stays under 50dB, which would make it far quieter than the rack servers MI300X cards normally live in.</p><p>Interestingly, these are not the MI300X modules AMD sells to hyperscalers. Tiny Corp refurbishes the cards itself, and they carry no warranty or support from AMD. Each sits on a carrier board Tiny Corp designed, with its own BMC microcontroller, rather than on a standard OAM module.</p><p>There are no XGMI links between the GPUs; each gets a dedicated PCIe 5.0 x16 link to the host instead. AMD's reference design gives every MI300X eight Infinity Fabric links at 128GB/s apiece, so in this box all GPU-to-GPU traffic travels a much narrower road. How that impacts performance is not yet known, with Tiny Corp offering no benchmarks.</p><p>Power is the other quiet compromise. The cards are capped at 500W via a 600W connector, compared to AMD's 750W board power rating. The 10.44 PFLOPS headline is AMD's peak at full power, so with each card running at two-thirds of that budget, sustained compute should theoretically land below the chipmaker's published figure.s </p><p>Tiny Corp hasn't published sustained numbers yet; however, memory bandwidth should suffer less from a power cap, and for single-user inference, that is the figure that matters most.</p><h2 id="a-great-deal-despite-tradeoffs">A great deal despite tradeoffs?</h2><p>The company is unusually candid about the tradeoffs. It warns that the software may be rough at first, but insists "none of the issues should be anything a good agent can't handle." </p><p>On X, it called the configuration plainly unsupported by AMD, while <a href="https://x.com/__tinygrad__/status/2118202833942286381" target="_blank" rel="nofollow">a companion post offered "huge thanks to AMD for supporting us"</a>. The two statements could run in parallel here if you read it as AMD having backed the company, not this particular build.</p><p>The listing also promises the cards will have no uncorrectable HBM errors. That guarantee matters more than it might seem: <a href="https://newsletter.semianalysis.com/p/can-amd-break-the-cuda-moat-amd-advancing" target="_blank" rel="nofollow">SemiAnalysis has reported</a> that Microsoft walked away from AMD after the MI300X, partly due to unreliable Samsung HBM from 2023. Tiny Corp hasn't said where its refurbished cards came from, so buyers are invoking a lot of trust here.</p><p>The MI300X box caps a remarkable reversal. In early 2024, Tiny Corp's first <a href="https://www.techradar.com/pro/another-startup-that-will-cause-gaming-gpu-prices-to-spike-ai-firm-claims-radeon-rx-7900-xtx-gpus-are-better-value-than-nvidias-h100-nearly-six-hundred-backers-believe-that-is-the-case" target="_blank">$15,000 tinybox with six Radeon RX 7900 XTX cards drew 583 preorders</a>. Within weeks, <a href="https://www.tomshardware.com/pc-components/gpus/amd-tinybox-project-put-on-hold-due-to-gpu-instability-in-ai-workloads-firm-publicly-considering-using-intel-gpus" target="_blank" rel="nofollow">the AMD version was put on hold</a> as the company publicly complained that the driver was still very unstable, before it <a href="https://www.phoronix.com/news/Tinybox-Radeon-Again-UMR" target="_blank" rel="nofollow">reversed course days later</a>. A year later, Hotz's <a href="https://geohot.github.io/blog/jekyll/update/2025/03/08/AMD-YOLO.html" target="_blank" rel="nofollow">"AMD YOLO" blog post</a> announced that AMD was shipping the two MI300X systems Tiny Corp had requested.</p><p>Tiny Corp also <a href="https://www.techradar.com/pro/tinygrad-refreshes-tinybox-with-an-old-epyc-cpu-running-ddr4-ram-to-feed-four-amd-9070xt-slaps-a-usd11-000-price-tag-on-a-new-computer-for-deep-learning" target="_blank">relaunched the tinybox in October around four Radeon RX 9070 XT cards for $11,000</a>, and weeks later <a href="https://www.techradar.com/pro/you-cant-currently-order-a-tinybox-with-nvidia-gpus-and-tiny-corps-new-192gb-six-gpu-amd-build-just-got-usd3-000-pricier" target="_blank">its configurator turned AMD-only after a commit titled "rip nvidia"</a>.</p><p>Despite the price hike, the secondhand market makes the box look cheap: Single used and refurbished MI300X OAM modules listed on eBay, as <a href="https://gpupoet.com/gpu/shop/amd-instinct-mi300x" target="_blank" rel="nofollow">tracked by GPU Poet</a>, were asking between about $27,550 and $40,300 on November 16, before anyone supplies a baseboard, host, or power. Even at $89,000, the tinybox works out to about $22,250 per GPU with the rest of the computer included, though asking prices on eBay are not sale prices.</p><p>Not everyone can buy one, either. The listing says the system is export-controlled and will only ship from San Diego to Tier 1 countries, meaning the US and its close allies. Orders are capped at two per customer, and buyers have 7 days to wire funds before an order is canceled.</p><p>A 48% price rise in under a week says plenty about demand for large pools of HBM outside the cloud. But batch two buyers are paying in 2026 for 2027 delivery of warranty-free, refurbished hardware in a configuration AMD doesn't support, and the tokens-per-second figure that would settle the value question has yet to be published officially by the company.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ AMD is adding tamper-resistant hardware to its chips to protect from physical attacks ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>AMD's Versal TR adaptive SoCs bond a sensor die to the silicon backside that detects physical attacks and blocks boot if removed</strong></li><li><strong>Post-quantum ML-DSA and SLH-DSA acceleration arrives after rivals Altera and Lattice already shipped post-quantum secure boot</strong></li><li><strong>Samples are not due until the second half of 2027, and AMD itself says tamper-resistant does not mean tamper-proof</strong></li></ul><p>AMD has unveiled a family of adaptive chips with their own built-in bodyguard.</p><p>The company's new <a href="https://newsroom.amd.com/news/amd-advances-built-in-protection-physical-attacks-security-critical-systems/" target="_blank" rel="nofollow">Versal tamper-resistant devices</a> bond a sensor-based protection die to the back of the main compute die, where it shields the silicon, watches for physical attacks, and stops the chip from booting if it is tampered with or removed.</p><p>The chips, specifically the 2VE3358TR, 2VE3558TR and 2VE3858TR, are all variants of the Versal AI Edge Series Gen 2 adaptive SoCs that AMD <a href="https://www.techradar.com/pro/amd-teams-up-with-arm-to-unveil-ai-chip-family-that-does-preprocessing-inference-and-postprocessing-on-one-silicon-but-you-will-have-to-wait-more-than-12-months-to-get-actual-products" target="_blank">first unveiled alongside Arm in </a><a href="https://www.techradar.com/pro/amd-teams-up-with-arm-to-unveil-ai-chip-family-that-does-preprocessing-inference-and-postprocessing-on-one-silicon-but-you-will-have-to-wait-more-than-12-months-to-get-actual-products" target="_blank">2024</a>, and will look to cater to the likes of defense systems, data center networking, financial technology and critical industrial infrastructure.</p><h2 id="a-second-die-guarding-the-side-that-nobody-watches">A second die guarding the side that nobody watches?</h2><p>One might ponder why the backside is suddenly important: most high-performance chips are flip-chip packaged, mounted face down, so the transistors sit against the package while the bare silicon substrate faces up.</p><p>Silicon is transparent to near-infrared light, so an attacker who removes the lid and thins the substrate can observe switching transistors through it, or fire a laser to inject faults, without ever disturbing the metal layers where traditional on-die mesh shields sit.</p><p><a href="https://www.electronics-journal.com/news/117307-tamper-resistant-adaptive-soc-protects-against-backside-physical-attacks" target="_blank" rel="nofollow">Electronics Journal notes</a> that defending this surface has generally relied on board- or enclosure-level measures, such as active meshes and potting. AMD knows this terrain from the receiving end: In September 2026, an AMD security bulletin covering its older 7-series FPGAs described a researcher recovering partial bitstream plaintext via contactless optical probing of the die's backside.</p><p>AMD's response was somewhat blunt then: that kind of access is <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-8034.html" target="_blank" rel="nofollow">"outside the threat model"</a> for those parts, and customers in high-risk deployments were advised to consider tamper-responsive enclosures, conformal coatings or potting.</p><p>It was not the first such case in recent times: AMD's <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-8022.html" target="_blank" rel="nofollow">February 2026 bulletin</a> on <a href="https://casa.rub.de/forschung/publikationen/detail/leakback-crc-optical-plaintext-recovery-of-encrypted-bitstreams-on-amd-7-series-fpgas" target="_blank" rel="nofollow">Leakback CRC</a>, a paper that reconstructed full netlists by optically probing the same family's readback CRC circuitry, drew the same out-of-scope verdict. The 7-series has history here, too: the <a href="https://www.techradar.com/news/starbleed-bug-threatens-fpga-chipsets-used-at-data-centers-iot-devices-around-the-world" target="_blank">Starbleed flaw disclosed in 2020</a> broke its bitstream encryption in a way researchers said could only be fixed by replacing the chip.</p><p>Versal TR's answer is to move that protection inside the package; The protection die is attached using what AMD calls 3D active-on-active technology, meaning both stacked dies carry active circuitry rather than one acting as a passive cap, and it covers the entire backside of the device.</p><p>AMD's <a href="https://www.amd.com/en/products/adaptive-socs-and-fpgas/tamper-resistant.html" target="_blank" rel="nofollow">product page for the parts</a> describes a full-die sensor array and "cryptographic binding" between the two dies. AMD has not explained that term, but it seems to imply that the shield is paired to its own compute die, so an attacker could not simply swap in a substitute.</p><p>AMD frames the pitch around consolidation; customers that need this level of protection today often build it in-house from enclosures, ASICs and other components, the company says, which adds cost, specialist engineering, verification work and the risk of a board re-spin.</p><p>Each TR part has a footprint-compatible standard counterpart that is available now, and the spec tables show identical logic, AI Engine and I/O resources on both, with backside protection and cryptographic binding as the only listed difference. Teams can develop on regular silicon and later swap in the TR device without rerunning synthesis or place and route.</p><p>AMD calls Versal TR the industry's first adaptive SoC or FPGA-based device with built-in protection against silicon backside attacks, a claim its footnote says rests on the company's own November 2026 analysis of competing parts.</p><p>In all fairness, backside shielding is not a new idea: CEA-Leti <a href="https://www.design-reuse.com/news/41587/leti-backside-shield-microchip-physical-attacks-protection.html" target="_blank" rel="nofollow">unveiled a backside shield for secure chips in 2017</a>, a metal serpentine sandwiched in polymer that typically triggers data erasure when disturbed. What has held such approaches back is cost. Researchers at Worcester Polytechnic Institute <a href="https://eprint.iacr.org/2024/631" target="_blank" rel="nofollow">wrote in 2024</a> that detection-based backside defenses are highly effective but expensive and highly customized. AMD's real contribution, if it delivers, is turning that into a catalog part.</p><p>The backside die might be the headline for most publications, but the TR parts also harden the front door. Versal already ships with tamper monitoring, and AMD's technical reference manual lists power-supply glitch detection, voltage and temperature alarms, and debug-port activity among its triggers.</p><h2 id="a-persistent-glitching-problem-as-mitigation-not-resolution-rule-the-roost">A persistent glitching problem as mitigation, not resolution rule the roost</h2><p>The TR devices improve the existing glitch detector and add a new orthogonal one that also catches glitches that hit two voltage rails at once. This is somewhat warranted because voltage glitching has a somewhat proven track record against AMD silicon. In 2021, TU Berlin researchers <a href="https://www.techradar.com/news/researchers-contradict-amd-claims-that-sev-keys-cant-be-extracted-remotely" target="_blank">used it to pull keys from the AMD Secure Processor</a> in Zen-based EPYC chips, and in 2023 a team from the same university <a href="https://www.techradar.com/news/some-of-the-top-amd-chips-are-suffering-a-serious-security-flaw" target="_blank">turned it on AMD's firmware TPM</a>. </p><p>After the latter, AMD said it was "continually innovating new hardware-based protections in future products." Yet when researchers reported extending the technique to a Zen 4 Ryzen 8600G in 2025, AMD's <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7044.html" target="_blank" rel="nofollow">bulletin</a> stated that physical attacks, such as voltage fault injection, fall outside the threat model for the affected products.</p><p>The approach AMD has applied here, however, makes the chips tamper-resistant rather than tamper-proof, and the chipmaker is extremely explicit about its limitations.  <a href="https://smbtech.au/news/amd-builds-physical-attack-defences-directly-into-new-versal-chip-family/" target="_blank" rel="nofollow">Its footnote states</a> that "tamper-resistant" does not mean "tamper-proof," that the protection applies only to specified physical attack methods, and that the threat model and limitations are set out in product documentation, which the product page does not publicly link to.</p><p>Versal TR also sits awkwardly beside AMD's stance on its server chips. When researchers broke SEV-SNP with a sub-$50 DDR4 interposer called Battering RAM in 2025, both AMD and Intel maintained that their trusted enclaves <a href="https://www.techradar.com/pro/security/battering-down-the-doors-usd50-hacking-kit-is-enough-to-break-intel-and-amds-toughest-chip-defenses" target="_blank">are not designed to resist physical attacks</a>, while the researchers said the work exposed "the fundamental limits of the scalable memory encryption designs" both companies use.</p><p>That position has held: in September 2026, AMD said a newer physical fault injection attack on DDR5 servers falls outside SEV-SNP's published threat model and that it <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-3048.html" target="_blank" rel="nofollow">"does not plan to assign a CVE or release mitigations."</a> Nothing in this announcement changes the EPYC threat model.</p><p>For the defense, fintech, and infrastructure customers AMD is courting, the appeal is plain: an off-the-shelf part that could replace a custom potted assembly, with a development path that starts on silicon available today.</p><p>Whether the protection die holds up against well-funded labs with failure analysis equipment is a question only independent testing can answer. That testing cannot begin until samples arrive in the second half of 2027, by which point competing Altera and Lattice customers will have had post-quantum secure boot for about a year.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/amd-is-adding-tamper-resistant-hardware-to-its-chips-to-protect-from-physical-attacks</link>
                                                                            <description>
                            <![CDATA[ AMD spent 2026 telling FPGA customers that backside probing was outside its threat model; now it is building backside protection into the silicon. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">9hvnh7MNqxa2okRBxbU2hh</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/7Wnzu6tGp6ew9yyA3nEQk9-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 21:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                <author><![CDATA[ Rahimnoorali11@gmail.com (Rahim Amir) ]]></author>                    <dc:creator><![CDATA[ Rahim Amir ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/9xKZFBamtEZKSChRvywbPB-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rahim Amir is a UAE-based tech writer who enjoys building PCs as much as he enjoys writing about them. He has been professionally writing about PC hardware since 2023, focusing on buyer’s guides, hardware reviews, and sponsored content and features related to tech.&lt;br&gt;&lt;br&gt;Having built hundreds of gaming PCs and being an avid gamer in his spare time, Rahim tends to have stronger opinions about hardware than most. This is particularly on display when he gets his way with powerful, but minimalistic RGB builds even as Small Form Factor (SFF) PCs come a close second.&lt;br&gt;&lt;br&gt;In addition to his contributions to TechRadar, Rahim’s work has also been featured on Game Rant and financial news websites.&lt;br&gt;&lt;br&gt;When he’s not working, you can find him playing DotA with friends or schmoozing to take the world over in Civilization. Alternatively, you can find him binging through the entirety of the Lord of The Rings universe with extended editions in play where applicable.&lt;br&gt;&lt;br&gt;You can currently catch Rahim grinding Path of Exile 2, complaining about his (extremely low) unique loot drop rate, or actively participating in one of the numerous (and heated) debates centered around Tolkien&#039;s universe on multiple forums daily.&lt;br&gt;&lt;br&gt;If you have a PC build or a Satisfactory playthrough in progress, he is likely to have some advice to send your way, especially regarding verticality being key for the latter. For the former, Rahim enjoys all aspects of the process including researching the components he will eventually use, benchmarking the latest and greatest hardware he can get his hands on, and somewhat surprisingly, cable management once he gets his latest build to POST.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/7Wnzu6tGp6ew9yyA3nEQk9-1920-80.png">
                                                            <media:credit><![CDATA[AMD]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[AMD Versal Gen 2 chip]]></media:description>                                                            <media:text><![CDATA[AMD Versal Gen 2 chip]]></media:text>
                                <media:title type="plain"><![CDATA[AMD Versal Gen 2 chip]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/7Wnzu6tGp6ew9yyA3nEQk9-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>AMD's Versal TR adaptive SoCs bond a sensor die to the silicon backside that detects physical attacks and blocks boot if removed</strong></li><li><strong>Post-quantum ML-DSA and SLH-DSA acceleration arrives after rivals Altera and Lattice already shipped post-quantum secure boot</strong></li><li><strong>Samples are not due until the second half of 2027, and AMD itself says tamper-resistant does not mean tamper-proof</strong></li></ul><p>AMD has unveiled a family of adaptive chips with their own built-in bodyguard.</p><p>The company's new <a href="https://newsroom.amd.com/news/amd-advances-built-in-protection-physical-attacks-security-critical-systems/" target="_blank" rel="nofollow">Versal tamper-resistant devices</a> bond a sensor-based protection die to the back of the main compute die, where it shields the silicon, watches for physical attacks, and stops the chip from booting if it is tampered with or removed.</p><p>The chips, specifically the 2VE3358TR, 2VE3558TR and 2VE3858TR, are all variants of the Versal AI Edge Series Gen 2 adaptive SoCs that AMD <a href="https://www.techradar.com/pro/amd-teams-up-with-arm-to-unveil-ai-chip-family-that-does-preprocessing-inference-and-postprocessing-on-one-silicon-but-you-will-have-to-wait-more-than-12-months-to-get-actual-products" target="_blank">first unveiled alongside Arm in </a><a href="https://www.techradar.com/pro/amd-teams-up-with-arm-to-unveil-ai-chip-family-that-does-preprocessing-inference-and-postprocessing-on-one-silicon-but-you-will-have-to-wait-more-than-12-months-to-get-actual-products" target="_blank">2024</a>, and will look to cater to the likes of defense systems, data center networking, financial technology and critical industrial infrastructure.</p><h2 id="a-second-die-guarding-the-side-that-nobody-watches">A second die guarding the side that nobody watches?</h2><p>One might ponder why the backside is suddenly important: most high-performance chips are flip-chip packaged, mounted face down, so the transistors sit against the package while the bare silicon substrate faces up.</p><p>Silicon is transparent to near-infrared light, so an attacker who removes the lid and thins the substrate can observe switching transistors through it, or fire a laser to inject faults, without ever disturbing the metal layers where traditional on-die mesh shields sit.</p><p><a href="https://www.electronics-journal.com/news/117307-tamper-resistant-adaptive-soc-protects-against-backside-physical-attacks" target="_blank" rel="nofollow">Electronics Journal notes</a> that defending this surface has generally relied on board- or enclosure-level measures, such as active meshes and potting. AMD knows this terrain from the receiving end: In September 2026, an AMD security bulletin covering its older 7-series FPGAs described a researcher recovering partial bitstream plaintext via contactless optical probing of the die's backside.</p><p>AMD's response was somewhat blunt then: that kind of access is <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-8034.html" target="_blank" rel="nofollow">"outside the threat model"</a> for those parts, and customers in high-risk deployments were advised to consider tamper-responsive enclosures, conformal coatings or potting.</p><p>It was not the first such case in recent times: AMD's <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-8022.html" target="_blank" rel="nofollow">February 2026 bulletin</a> on <a href="https://casa.rub.de/forschung/publikationen/detail/leakback-crc-optical-plaintext-recovery-of-encrypted-bitstreams-on-amd-7-series-fpgas" target="_blank" rel="nofollow">Leakback CRC</a>, a paper that reconstructed full netlists by optically probing the same family's readback CRC circuitry, drew the same out-of-scope verdict. The 7-series has history here, too: the <a href="https://www.techradar.com/news/starbleed-bug-threatens-fpga-chipsets-used-at-data-centers-iot-devices-around-the-world" target="_blank">Starbleed flaw disclosed in 2020</a> broke its bitstream encryption in a way researchers said could only be fixed by replacing the chip.</p><p>Versal TR's answer is to move that protection inside the package; The protection die is attached using what AMD calls 3D active-on-active technology, meaning both stacked dies carry active circuitry rather than one acting as a passive cap, and it covers the entire backside of the device.</p><p>AMD's <a href="https://www.amd.com/en/products/adaptive-socs-and-fpgas/tamper-resistant.html" target="_blank" rel="nofollow">product page for the parts</a> describes a full-die sensor array and "cryptographic binding" between the two dies. AMD has not explained that term, but it seems to imply that the shield is paired to its own compute die, so an attacker could not simply swap in a substitute.</p><p>AMD frames the pitch around consolidation; customers that need this level of protection today often build it in-house from enclosures, ASICs and other components, the company says, which adds cost, specialist engineering, verification work and the risk of a board re-spin.</p><p>Each TR part has a footprint-compatible standard counterpart that is available now, and the spec tables show identical logic, AI Engine and I/O resources on both, with backside protection and cryptographic binding as the only listed difference. Teams can develop on regular silicon and later swap in the TR device without rerunning synthesis or place and route.</p><p>AMD calls Versal TR the industry's first adaptive SoC or FPGA-based device with built-in protection against silicon backside attacks, a claim its footnote says rests on the company's own November 2026 analysis of competing parts.</p><p>In all fairness, backside shielding is not a new idea: CEA-Leti <a href="https://www.design-reuse.com/news/41587/leti-backside-shield-microchip-physical-attacks-protection.html" target="_blank" rel="nofollow">unveiled a backside shield for secure chips in 2017</a>, a metal serpentine sandwiched in polymer that typically triggers data erasure when disturbed. What has held such approaches back is cost. Researchers at Worcester Polytechnic Institute <a href="https://eprint.iacr.org/2024/631" target="_blank" rel="nofollow">wrote in 2024</a> that detection-based backside defenses are highly effective but expensive and highly customized. AMD's real contribution, if it delivers, is turning that into a catalog part.</p><p>The backside die might be the headline for most publications, but the TR parts also harden the front door. Versal already ships with tamper monitoring, and AMD's technical reference manual lists power-supply glitch detection, voltage and temperature alarms, and debug-port activity among its triggers.</p><h2 id="a-persistent-glitching-problem-as-mitigation-not-resolution-rule-the-roost">A persistent glitching problem as mitigation, not resolution rule the roost</h2><p>The TR devices improve the existing glitch detector and add a new orthogonal one that also catches glitches that hit two voltage rails at once. This is somewhat warranted because voltage glitching has a somewhat proven track record against AMD silicon. In 2021, TU Berlin researchers <a href="https://www.techradar.com/news/researchers-contradict-amd-claims-that-sev-keys-cant-be-extracted-remotely" target="_blank">used it to pull keys from the AMD Secure Processor</a> in Zen-based EPYC chips, and in 2023 a team from the same university <a href="https://www.techradar.com/news/some-of-the-top-amd-chips-are-suffering-a-serious-security-flaw" target="_blank">turned it on AMD's firmware TPM</a>. </p><p>After the latter, AMD said it was "continually innovating new hardware-based protections in future products." Yet when researchers reported extending the technique to a Zen 4 Ryzen 8600G in 2025, AMD's <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-7044.html" target="_blank" rel="nofollow">bulletin</a> stated that physical attacks, such as voltage fault injection, fall outside the threat model for the affected products.</p><p>The approach AMD has applied here, however, makes the chips tamper-resistant rather than tamper-proof, and the chipmaker is extremely explicit about its limitations.  <a href="https://smbtech.au/news/amd-builds-physical-attack-defences-directly-into-new-versal-chip-family/" target="_blank" rel="nofollow">Its footnote states</a> that "tamper-resistant" does not mean "tamper-proof," that the protection applies only to specified physical attack methods, and that the threat model and limitations are set out in product documentation, which the product page does not publicly link to.</p><p>Versal TR also sits awkwardly beside AMD's stance on its server chips. When researchers broke SEV-SNP with a sub-$50 DDR4 interposer called Battering RAM in 2025, both AMD and Intel maintained that their trusted enclaves <a href="https://www.techradar.com/pro/security/battering-down-the-doors-usd50-hacking-kit-is-enough-to-break-intel-and-amds-toughest-chip-defenses" target="_blank">are not designed to resist physical attacks</a>, while the researchers said the work exposed "the fundamental limits of the scalable memory encryption designs" both companies use.</p><p>That position has held: in September 2026, AMD said a newer physical fault injection attack on DDR5 servers falls outside SEV-SNP's published threat model and that it <a href="https://www.amd.com/en/resources/product-security/bulletin/amd-sb-3048.html" target="_blank" rel="nofollow">"does not plan to assign a CVE or release mitigations."</a> Nothing in this announcement changes the EPYC threat model.</p><p>For the defense, fintech, and infrastructure customers AMD is courting, the appeal is plain: an off-the-shelf part that could replace a custom potted assembly, with a development path that starts on silicon available today.</p><p>Whether the protection die holds up against well-funded labs with failure analysis equipment is a question only independent testing can answer. That testing cannot begin until samples arrive in the second half of 2027, by which point competing Altera and Lattice customers will have had post-quantum secure boot for about a year.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Kioxia and Dell have managed to cram in a huge 157 PB of raw flash storage into a single 42U server rack ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Dell fits 640 high-capacity SSDs into a standard 42U server rack</strong></li><li><strong>Kioxia's latest configuration delivers 157 PB across just 16 servers</strong></li><li><strong>The Dell installation occupies 32U, leaving 10U available for equipment</strong></li></ul><p>Kioxia and Dell Technologies have combined 640 high-capacity solid-state drives to achieve 157 PB of raw flash storage within one server rack.</p><p>The configuration uses 16 Dell PowerEdge R7725xd servers, each equipped with 40 Kioxia LC9 Series <a href="https://www.techradar.com/news/best-solid-state-drives-ssds">SSDs</a> rated at 245.76 TB.</p><p>The companies <a href="https://www.businesswire.com/news/home/20261111365949/en/KIOXIA-and-Dell-Technologies-Collaborate-on-the-Industrys-First-157-PB-Single-Rack-Object-Store-Redefining-Storage-Economics" target="_blank" rel="nofollow">describe</a> the installation as an industry-first achievement, although competing storage developers have reported configurations with even greater rack capacities.</p><h2 id="640-ssds-deliver-157-pb-of-storage-capacity">640 SSDs deliver 157 PB of storage capacity</h2><p>The installation runs Dell's ObjectScale software, an object-storage platform designed to manage enormous collections of unstructured information.</p><p>Its intended applications include artificial intelligence training, large-scale data analysis, and cloud services requiring substantial storage capacity.</p><p>Each PowerEdge server occupies 2U of rack space and accommodates 40 E3.L SSDs - and with 16 servers installed, the configuration consumes 32U of the standard 42U enclosure, leaving another 10U available for additional equipment.</p><p>The arrangement follows an earlier collaboration between the companies, which produced a configuration offering 9.8 PB of flash storage per server.</p><p>Multiplying that capacity across 16 machines produces approximately 157 PB, explaining how the latest installation reaches its advertised figure.</p><p>Kioxia's LC9 drives use eighth-generation BiCS FLASH technology, incorporating quad-level cell architecture and a 32-die stacking arrangement.</p><p>The drives also support PCIe 5.0 and NVMe 2.0, while their advertised mean time to failure reaches 2.5 million hours.</p><p>However, the maximum 245.76 TB capacity applies specifically to the E3.L version. The alternative 2.5-inch and E3.S models offer capacities reaching 122.88 TB, substantially reducing the available storage within an equivalent configuration.</p><h2 id="power-consumption-and-competing-drives-remain-important-considerations">Power consumption and competing drives remain important considerations</h2><p>The companies argue that concentrating more information into fewer servers can reduce cabling requirements, electricity consumption and cooling demands.</p><p>Dell also believes the arrangement could help organizations accommodate expanding AI datasets without continually increasing their physical storage infrastructure.</p><p>“AI is driving unprecedented growth in unstructured data, and customers need to scale capacity without scaling their <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> footprint and power consumption,” said George Papadopoulos, SVP, Product &amp; Engineering, Dell Technologies.</p><p>“With Dell ObjectScale and KIOXIA’s 245.76 TB SSDs, customers can consolidate more data into fewer systems and racks while reducing cabling, power and cooling requirements. This collaboration helps deliver a more efficient, scalable foundation for AI and data-intensive workloads.”</p><p>Nevertheless, the announcement provides no measured electricity consumption, cooling figures or comparative performance results establishing the actual savings.</p><p>It also omits pricing and availability information, making the financial advantages difficult to assess.</p><p>Competition is another consideration, particularly because Micron has introduced its own 245 TB 6600 ION SSD.</p><p>The company advertises configurations reaching 176.9 PB within a rack using the E3.L form factor, exceeding Kioxia's stated capacity.</p><p><a href="https://www.techradar.com/pro/kiwi-storage-startup-uses-proprietary-144tb-ssd-to-hit-230tb-storage-per-42u-rack-but-it-will-cost-well-over-usd2-million-to-get-such-a-system">Novodisq has also reported an even higher capacity</a>, claiming that its proprietary 144 TB SSDs can deliver 230 PB within a standard 42U rack.</p><p>Its Novoblade architecture uses modular storage blades containing multiple drives, allowing the company to achieve approximately 46% more raw capacity than the Kioxia and Dell installation.</p><p>However, independently verified performance, power consumption and cooling measurements would help establish whether its architecture delivers comparable operational advantages.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/kioxia-and-dell-have-managed-to-cram-in-a-huge-157-pb-of-raw-flash-storage-into-a-single-42u-server-rack</link>
                                                                            <description>
                            <![CDATA[ Kioxia and Dell combine 640 high-capacity SSDs across 16 servers, delivering 157 PB of raw flash storage within one standard rack. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">4AJJDwnfyNJYFyTyRYou2D</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/hbYdzu72tW2nnyfp6EDPcR-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 20:40:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/hbYdzu72tW2nnyfp6EDPcR-1920-80.png">
                                                            <media:credit><![CDATA[Kioxia]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Kioxia LC9 BiCS FLASH]]></media:description>                                                            <media:text><![CDATA[Kioxia LC9 BiCS FLASH]]></media:text>
                                <media:title type="plain"><![CDATA[Kioxia LC9 BiCS FLASH]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/hbYdzu72tW2nnyfp6EDPcR-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Dell fits 640 high-capacity SSDs into a standard 42U server rack</strong></li><li><strong>Kioxia's latest configuration delivers 157 PB across just 16 servers</strong></li><li><strong>The Dell installation occupies 32U, leaving 10U available for equipment</strong></li></ul><p>Kioxia and Dell Technologies have combined 640 high-capacity solid-state drives to achieve 157 PB of raw flash storage within one server rack.</p><p>The configuration uses 16 Dell PowerEdge R7725xd servers, each equipped with 40 Kioxia LC9 Series <a href="https://www.techradar.com/news/best-solid-state-drives-ssds">SSDs</a> rated at 245.76 TB.</p><p>The companies <a href="https://www.businesswire.com/news/home/20261111365949/en/KIOXIA-and-Dell-Technologies-Collaborate-on-the-Industrys-First-157-PB-Single-Rack-Object-Store-Redefining-Storage-Economics" target="_blank" rel="nofollow">describe</a> the installation as an industry-first achievement, although competing storage developers have reported configurations with even greater rack capacities.</p><h2 id="640-ssds-deliver-157-pb-of-storage-capacity">640 SSDs deliver 157 PB of storage capacity</h2><p>The installation runs Dell's ObjectScale software, an object-storage platform designed to manage enormous collections of unstructured information.</p><p>Its intended applications include artificial intelligence training, large-scale data analysis, and cloud services requiring substantial storage capacity.</p><p>Each PowerEdge server occupies 2U of rack space and accommodates 40 E3.L SSDs - and with 16 servers installed, the configuration consumes 32U of the standard 42U enclosure, leaving another 10U available for additional equipment.</p><p>The arrangement follows an earlier collaboration between the companies, which produced a configuration offering 9.8 PB of flash storage per server.</p><p>Multiplying that capacity across 16 machines produces approximately 157 PB, explaining how the latest installation reaches its advertised figure.</p><p>Kioxia's LC9 drives use eighth-generation BiCS FLASH technology, incorporating quad-level cell architecture and a 32-die stacking arrangement.</p><p>The drives also support PCIe 5.0 and NVMe 2.0, while their advertised mean time to failure reaches 2.5 million hours.</p><p>However, the maximum 245.76 TB capacity applies specifically to the E3.L version. The alternative 2.5-inch and E3.S models offer capacities reaching 122.88 TB, substantially reducing the available storage within an equivalent configuration.</p><h2 id="power-consumption-and-competing-drives-remain-important-considerations">Power consumption and competing drives remain important considerations</h2><p>The companies argue that concentrating more information into fewer servers can reduce cabling requirements, electricity consumption and cooling demands.</p><p>Dell also believes the arrangement could help organizations accommodate expanding AI datasets without continually increasing their physical storage infrastructure.</p><p>“AI is driving unprecedented growth in unstructured data, and customers need to scale capacity without scaling their <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> footprint and power consumption,” said George Papadopoulos, SVP, Product &amp; Engineering, Dell Technologies.</p><p>“With Dell ObjectScale and KIOXIA’s 245.76 TB SSDs, customers can consolidate more data into fewer systems and racks while reducing cabling, power and cooling requirements. This collaboration helps deliver a more efficient, scalable foundation for AI and data-intensive workloads.”</p><p>Nevertheless, the announcement provides no measured electricity consumption, cooling figures or comparative performance results establishing the actual savings.</p><p>It also omits pricing and availability information, making the financial advantages difficult to assess.</p><p>Competition is another consideration, particularly because Micron has introduced its own 245 TB 6600 ION SSD.</p><p>The company advertises configurations reaching 176.9 PB within a rack using the E3.L form factor, exceeding Kioxia's stated capacity.</p><p><a href="https://www.techradar.com/pro/kiwi-storage-startup-uses-proprietary-144tb-ssd-to-hit-230tb-storage-per-42u-rack-but-it-will-cost-well-over-usd2-million-to-get-such-a-system">Novodisq has also reported an even higher capacity</a>, claiming that its proprietary 144 TB SSDs can deliver 230 PB within a standard 42U rack.</p><p>Its Novoblade architecture uses modular storage blades containing multiple drives, allowing the company to achieve approximately 46% more raw capacity than the Kioxia and Dell installation.</p><p>However, independently verified performance, power consumption and cooling measurements would help establish whether its architecture delivers comparable operational advantages.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Hackers are using security cameras to break into corporate networks, then switching off the antivirus to mine crypto ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Hackers exploited exposed GeoVision surveillance systems to gain access to corporate networks</strong></li><li><strong>Attackers manipulated Windows settings to interfere with Microsoft Defender protections</strong></li><li><strong>Compromised computers were turned into cryptocurrency mining machines using XMRig</strong></li></ul><p>Hackers have exploited internet-accessible surveillance systems to infiltrate corporate networks, disable security protections, and install cryptocurrency mining software on compromised computers.</p><p>The attacks abused GeoVision webcam software, allowing intruders to execute malicious commands through exposed services and gain control over affected Windows machines.</p><p>Cybersecurity researchers at <a href="https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer" target="_blank" rel="nofollow">Huntress</a> identified the activity across 13 organizations, with infections affecting ten accounts between November 8 and November 10 2026.</p><h2 id="exposed-surveillance-systems-provide-an-entry-point-for-attackers">Exposed surveillance systems provide an entry point for attackers</h2><p>The investigation found that attackers exploited publicly accessible GeoVision devices and management interfaces to gain access to connected organizational networks.</p><p>Once inside, malicious commands originating from GeoStartupAgentService.exe launched Windows command shells that subsequently executed downloaded PowerShell scripts.</p><p>These scripts examined infected computers, checked whether Microsoft Defender was running, and searched for writable directories before introducing additional malicious components.</p><p>One downloaded component, kvckiller.sys, was identified as a vulnerable Windows kernel driver previously abused to interfere with <a href="https://www.techradar.com/best/best-antivirus">antivirus</a> protections.</p><p>The attackers also manipulated Windows Registry settings to prevent essential security applications from launching through their normal executable files.</p><p>Three Microsoft Defender and Windows Security components were affected, including MsMpEng.exe, SecurityHealthService.exe, and SecurityHealthSystray.exe.</p><p>The changes redirected attempts to launch these applications toward systray.exe, potentially preventing their intended security functions from operating normally.</p><p>This gave the intruders another opportunity to maintain access while reducing the likelihood that conventional antivirus protections would detect malicious activity.</p><p>Huntress also discovered that attackers used scheduled tasks to ensure their malicious programs continued running after initial execution or system restarts.</p><p>One task, PackageUpdater, instructed Windows to execute a malicious program called update.exe every 15 minutes, restoring it whenever necessary.</p><p>Another scheduled task, PackageUpdater_Startup, was configured to launch the same program automatically whenever an affected computer started. </p><h2 id="xmrig-turns-compromised-computers-into-cryptocurrency-mining-machines">XMRig turns compromised computers into cryptocurrency mining machines</h2><p>After disabling security protections, the attackers deployed XMRig, an open-source cryptocurrency mining program, through a custom loader written in Golang.</p><p>The malicious loader, named update.exe, downloaded another executable called ApplyUpdate.exe, which contained the cryptocurrency mining software.</p><p>XMRig uses computer processing resources to perform cryptocurrency mining operations, allowing attackers to generate revenue using machines belonging to unsuspecting organizations.</p><p>The loader also incorporated BitTorrent Mainline Distributed Hash Table technology and a PKARR resolver to locate its command infrastructure.</p><p>This arrangement allowed the operators to discover changing server addresses without depending entirely on conventional domain registration and fixed infrastructure.</p><p>Huntress identified several indicators associated with the campaign, including suspicious connections involving WebCamServer.exe and unusual DNS requests from compromised systems.</p><p>Investigators also identified malicious infrastructure associated with the activity, including the domain cctv-update.top and several IP addresses used during different attack stages.</p><p>The researchers advised organizations to examine GeoVision installations for unexpected command shells, PowerShell activity, and suspicious scheduled tasks.</p><p>They also recommended removing unnecessary public access to surveillance equipment, using <a href="https://www.techradar.com/vpn/best-vpn">VPN</a> connections, and restricting management interfaces to trusted addresses.</p><p>For organizations running these systems, keeping camera management software outside unrestricted internet access could reduce opportunities for attackers to exploit exposed services.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/hackers-are-using-security-cameras-to-break-into-corporate-networks-then-switching-off-the-antivirus-to-mine-crypto</link>
                                                                            <description>
                            <![CDATA[ Hackers exploited exposed GeoVision surveillance systems to infiltrate corporate networks, interfere with antivirus protections, and install cryptocurrency mining software on computers. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">2i6EpHpY8uXY4kRXHhPBpg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/yf4vqPknb8WxxxzfBvhVjZ-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 19:35:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Cyber Crime]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/yf4vqPknb8WxxxzfBvhVjZ-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Hackers using AI]]></media:description>                                                            <media:text><![CDATA[Hackers using AI]]></media:text>
                                <media:title type="plain"><![CDATA[Hackers using AI]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/yf4vqPknb8WxxxzfBvhVjZ-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Hackers exploited exposed GeoVision surveillance systems to gain access to corporate networks</strong></li><li><strong>Attackers manipulated Windows settings to interfere with Microsoft Defender protections</strong></li><li><strong>Compromised computers were turned into cryptocurrency mining machines using XMRig</strong></li></ul><p>Hackers have exploited internet-accessible surveillance systems to infiltrate corporate networks, disable security protections, and install cryptocurrency mining software on compromised computers.</p><p>The attacks abused GeoVision webcam software, allowing intruders to execute malicious commands through exposed services and gain control over affected Windows machines.</p><p>Cybersecurity researchers at <a href="https://www.huntress.com/blog/geovision-camera-software-xmrig-cryptominer" target="_blank" rel="nofollow">Huntress</a> identified the activity across 13 organizations, with infections affecting ten accounts between November 8 and November 10 2026.</p><h2 id="exposed-surveillance-systems-provide-an-entry-point-for-attackers">Exposed surveillance systems provide an entry point for attackers</h2><p>The investigation found that attackers exploited publicly accessible GeoVision devices and management interfaces to gain access to connected organizational networks.</p><p>Once inside, malicious commands originating from GeoStartupAgentService.exe launched Windows command shells that subsequently executed downloaded PowerShell scripts.</p><p>These scripts examined infected computers, checked whether Microsoft Defender was running, and searched for writable directories before introducing additional malicious components.</p><p>One downloaded component, kvckiller.sys, was identified as a vulnerable Windows kernel driver previously abused to interfere with <a href="https://www.techradar.com/best/best-antivirus">antivirus</a> protections.</p><p>The attackers also manipulated Windows Registry settings to prevent essential security applications from launching through their normal executable files.</p><p>Three Microsoft Defender and Windows Security components were affected, including MsMpEng.exe, SecurityHealthService.exe, and SecurityHealthSystray.exe.</p><p>The changes redirected attempts to launch these applications toward systray.exe, potentially preventing their intended security functions from operating normally.</p><p>This gave the intruders another opportunity to maintain access while reducing the likelihood that conventional antivirus protections would detect malicious activity.</p><p>Huntress also discovered that attackers used scheduled tasks to ensure their malicious programs continued running after initial execution or system restarts.</p><p>One task, PackageUpdater, instructed Windows to execute a malicious program called update.exe every 15 minutes, restoring it whenever necessary.</p><p>Another scheduled task, PackageUpdater_Startup, was configured to launch the same program automatically whenever an affected computer started. </p><h2 id="xmrig-turns-compromised-computers-into-cryptocurrency-mining-machines">XMRig turns compromised computers into cryptocurrency mining machines</h2><p>After disabling security protections, the attackers deployed XMRig, an open-source cryptocurrency mining program, through a custom loader written in Golang.</p><p>The malicious loader, named update.exe, downloaded another executable called ApplyUpdate.exe, which contained the cryptocurrency mining software.</p><p>XMRig uses computer processing resources to perform cryptocurrency mining operations, allowing attackers to generate revenue using machines belonging to unsuspecting organizations.</p><p>The loader also incorporated BitTorrent Mainline Distributed Hash Table technology and a PKARR resolver to locate its command infrastructure.</p><p>This arrangement allowed the operators to discover changing server addresses without depending entirely on conventional domain registration and fixed infrastructure.</p><p>Huntress identified several indicators associated with the campaign, including suspicious connections involving WebCamServer.exe and unusual DNS requests from compromised systems.</p><p>Investigators also identified malicious infrastructure associated with the activity, including the domain cctv-update.top and several IP addresses used during different attack stages.</p><p>The researchers advised organizations to examine GeoVision installations for unexpected command shells, PowerShell activity, and suspicious scheduled tasks.</p><p>They also recommended removing unnecessary public access to surveillance equipment, using <a href="https://www.techradar.com/vpn/best-vpn">VPN</a> connections, and restricting management interfaces to trusted addresses.</p><p>For organizations running these systems, keeping camera management software outside unrestricted internet access could reduce opportunities for attackers to exploit exposed services.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Uber data breach sees hackers steal license information on thousands of drivers ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Uber discloses third-party breach that exposed driver data, including driver’s license information for thousands of individuals</strong></li><li><strong>The incident stemmed from a compromise at AI infrastructure provider Fireworks, which was also targeted in broader LLMjacking attacks</strong></li><li><strong>Uber is notifying affected users and offering identity theft and credit monitoring services while investigations continue</strong></li></ul><p>Popular ride-hailing service Uber has confirmed it lost sensitive data after one of its third-party service providers was hacked. </p><p>The company filed a new notice with multiple US Attorney General’s Offices and shared a copy of the letter it is now sending out to affected individuals. </p><p>In the letter, it says that a threat actor got ahold of <a href="https://www.techradar.com/best/password-manager" target="_blank">login credentials</a> belonging to an employee of the service and used them to pull Uber’s data.</p><h2 id="driver-s-licenses-and-other-data">Driver’s licenses and other data</h2><p>The letter does not disclose key details, like the nature of the data that was stolen, or how many people were affected. </p><p>According to<em> </em><a href="https://techcrunch.com/2026/11/13/uber-confirms-hackers-stole-thousands-of-drivers-license-information-in-data-breach/" target="_blank"><em>TechCrunch</em></a>, however, “thousands” of drivers lost personal and driver’s license information, with the stolen data including driver’s licenses of 2,567 individuals in Massachusetts, 3,430 people across South Carolina, and 1,873 residents of Rhode Island.</p><p>More drivers in Illinois and California, and possibly elsewhere, may also be affected - but these figures are so far unconfirmed. </p><p>While Uber did not say which of its third-party service providers got hacked, the media uncovered it was a company called Fireworks AI which, itself, was also a victim of a third-party supply chain attack. Fireworks is a US-based AI infrastructure company that helps businesses deploy and run AI models at scale. Apparently, it provides access to hundreds of AI models, as well as the computing resources needed to run them, and its clients include heavy-hitters such as Shopify, DoorDash, and Notion. </p><p>Fireworks recently issued a notice saying it spotted a threat actor using internal company credentials to access “environment variables used in evaluation jobs” in a “limited number” of customer accounts. </p><p>At first, the company said it did not find evidence of unauthorized access to customer models, datasets, training data, personal information, or inference traffic, but a subsequent update changed the wording somewhat: “At the time of publishing the bulletin, only Fireworks stored secrets (e.g. used in evaluation jobs) were known to have been accessed. Since then, we discovered impacts to account and billing information. For a limited set of accounts, we also discovered impact to secrets, datasets, and incomplete models,” it said.</p><h2 id="llmjacking">LLMjacking</h2><p>The data stolen from Uber seems like added value, not the primary target, as Fireworks says all evidence indicates this was an <a href="https://www.techradar.com/pro/security/hackers-are-using-llmjacking-attacks-to-hijack-company-ai-usage-and-run-up-huge-cloud-bills" target="_blank">LLMjacking attack</a>.</p><p>LLMjacking is a relatively new term. Google’s Threat Intelligence Group (GTIG) recently described it as “the monetization and resource-hijacking of foundation models, which actors achieve through token proxying, automated model streaming, and GPU hardware provisioning.” In other words, someone else is using your AI subscription while you get the bill.</p><p>The attacks are quite disruptive, too. Unlike cryptojacking, which takes a long time to rake up any meaningful bills, LLMjacking costs can go up in mere minutes. Google’s recent report showed how a Chinese actor tracked as UNC6993 cost its victims more than $100,000. </p><p>The company suffered a secondary attack, as well, less than a month later. On October 23, someone broke into a third-party cloud provider that hosted the company’s R&amp;D machines and used it to reach Fireworks development machines. </p><p>There, they ran unauthorized workloads “that led to additional customer data impact.” The attack was spotted on the same day and stopped. “As of October 25, we have not detected further evidence of successful unauthorized access by the threat actor on the Fireworks platform customers use,” the company concluded. </p><p>Both Fireworks and Uber are now notifying affected individuals of the breach, and the latter is also offering free identity theft and credit monitoring through Cyberscout. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/uber-data-breach-sees-hackers-steal-license-information-on-thousands-of-drivers</link>
                                                                            <description>
                            <![CDATA[ One of Uber's services was breached, and its data stolen. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">gEHugin63GTAtq9wmWZCiM</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/aFuqJoM5FFuvC5JASPq2TG-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 18:15:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/aFuqJoM5FFuvC5JASPq2TG-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock / Sundry Photography]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Close up of UBER sign at their headquarters in San Francisco&#039;s SOMA district]]></media:description>                                                            <media:text><![CDATA[Close up of UBER sign at their headquarters in San Francisco&#039;s SOMA district]]></media:text>
                                <media:title type="plain"><![CDATA[Close up of UBER sign at their headquarters in San Francisco&#039;s SOMA district]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/aFuqJoM5FFuvC5JASPq2TG-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Uber discloses third-party breach that exposed driver data, including driver’s license information for thousands of individuals</strong></li><li><strong>The incident stemmed from a compromise at AI infrastructure provider Fireworks, which was also targeted in broader LLMjacking attacks</strong></li><li><strong>Uber is notifying affected users and offering identity theft and credit monitoring services while investigations continue</strong></li></ul><p>Popular ride-hailing service Uber has confirmed it lost sensitive data after one of its third-party service providers was hacked. </p><p>The company filed a new notice with multiple US Attorney General’s Offices and shared a copy of the letter it is now sending out to affected individuals. </p><p>In the letter, it says that a threat actor got ahold of <a href="https://www.techradar.com/best/password-manager" target="_blank">login credentials</a> belonging to an employee of the service and used them to pull Uber’s data.</p><h2 id="driver-s-licenses-and-other-data">Driver’s licenses and other data</h2><p>The letter does not disclose key details, like the nature of the data that was stolen, or how many people were affected. </p><p>According to<em> </em><a href="https://techcrunch.com/2026/11/13/uber-confirms-hackers-stole-thousands-of-drivers-license-information-in-data-breach/" target="_blank"><em>TechCrunch</em></a>, however, “thousands” of drivers lost personal and driver’s license information, with the stolen data including driver’s licenses of 2,567 individuals in Massachusetts, 3,430 people across South Carolina, and 1,873 residents of Rhode Island.</p><p>More drivers in Illinois and California, and possibly elsewhere, may also be affected - but these figures are so far unconfirmed. </p><p>While Uber did not say which of its third-party service providers got hacked, the media uncovered it was a company called Fireworks AI which, itself, was also a victim of a third-party supply chain attack. Fireworks is a US-based AI infrastructure company that helps businesses deploy and run AI models at scale. Apparently, it provides access to hundreds of AI models, as well as the computing resources needed to run them, and its clients include heavy-hitters such as Shopify, DoorDash, and Notion. </p><p>Fireworks recently issued a notice saying it spotted a threat actor using internal company credentials to access “environment variables used in evaluation jobs” in a “limited number” of customer accounts. </p><p>At first, the company said it did not find evidence of unauthorized access to customer models, datasets, training data, personal information, or inference traffic, but a subsequent update changed the wording somewhat: “At the time of publishing the bulletin, only Fireworks stored secrets (e.g. used in evaluation jobs) were known to have been accessed. Since then, we discovered impacts to account and billing information. For a limited set of accounts, we also discovered impact to secrets, datasets, and incomplete models,” it said.</p><h2 id="llmjacking">LLMjacking</h2><p>The data stolen from Uber seems like added value, not the primary target, as Fireworks says all evidence indicates this was an <a href="https://www.techradar.com/pro/security/hackers-are-using-llmjacking-attacks-to-hijack-company-ai-usage-and-run-up-huge-cloud-bills" target="_blank">LLMjacking attack</a>.</p><p>LLMjacking is a relatively new term. Google’s Threat Intelligence Group (GTIG) recently described it as “the monetization and resource-hijacking of foundation models, which actors achieve through token proxying, automated model streaming, and GPU hardware provisioning.” In other words, someone else is using your AI subscription while you get the bill.</p><p>The attacks are quite disruptive, too. Unlike cryptojacking, which takes a long time to rake up any meaningful bills, LLMjacking costs can go up in mere minutes. Google’s recent report showed how a Chinese actor tracked as UNC6993 cost its victims more than $100,000. </p><p>The company suffered a secondary attack, as well, less than a month later. On October 23, someone broke into a third-party cloud provider that hosted the company’s R&amp;D machines and used it to reach Fireworks development machines. </p><p>There, they ran unauthorized workloads “that led to additional customer data impact.” The attack was spotted on the same day and stopped. “As of October 25, we have not detected further evidence of successful unauthorized access by the threat actor on the Fireworks platform customers use,” the company concluded. </p><p>Both Fireworks and Uber are now notifying affected individuals of the breach, and the latter is also offering free identity theft and credit monitoring through Cyberscout. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ TSMC confirms cyberattack and data breach — chip giant says personal employee data has been leaked online ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>TSMC confirmed employee data was exposed through a third-party cloud provider rather than its own systems</strong></li><li><strong>Stolen records reportedly include employee personal and banking information, with around 46,000 records offered for sale</strong></li><li><strong>Company is notifying affected personnel and investigating the breach alongside authorities and security experts</strong></li></ul><p>TSMC, the world’s largest semiconductor manufacturer, has confirmed suffering a cyberattack and losing sensitive files after a threat actor offered a large database for sale on the dark web.</p><p>This comes after a threat actor with the alias “SmithBianl” recently posted a new thread on an underground forum offering a text file with 499 rows of data they claim was nabbed from TSMC. The text file, they claim, provides a sample of a much larger dataset they owned, which contains 46,000 employee records and weighs almost 7GB.</p><p>They also said the stolen archive contains TSMC employees’ national ID numbers, addresses, contact details, and banking information, and are asking 3 BTC in exchange (just below $200,000 at current prices). They also said that they would leak the entire archive if a buyer doesn’t step forward by November 15 2026 - a deadline which has now expired. </p><h2 id="third-party-cloud-service-provider-struck">Third-party cloud service provider struck</h2><p>Following the leak, the company’s Senior Vice President and Chief Financial Officer (CFO), Wendell Huang, issued a short statement confirming that the company was not breached and that the stolen files were hosted on a third party:</p><p>“TSMC has found that some of our current and former employee data was made available on an external website without authorization,” the announcement read.</p><p>“Currently, we have found no evidence that TSMC’s systems have been compromised. This information has been verified to match data stored in the system services of a cloud service provider, but is unrelated to the company’s operations or customer-related data.”</p><p>Huang would not say which third-party <a href="https://www.techradar.com/best/best-cloud-storage" target="_blank">cloud provider</a> that was, and the company is currently not responding to media inquiries. Therefore, we also don’t know how that company was breached, and if it hosted - and lost - data belonging to other companies, as well. </p><p>We also don’t know exactly how many people had their data exposed in this incident. 46,000 employee records doesn’t necessarily mean 46,000 people, since victims lost different records such as names and banking information.</p><p>Some of them might be TSMC staff, given the curious choice of words in Huang’s letter. </p><p>“We have been informing impacted personnel and are providing them with necessary support,” he said. “TSMC has engaged with both the cloud service vendor, the relevant investigating authorities, and a third-party cybersecurity agency to jointly ascertain the facts and take necessary actions. We remain firmly committed to protecting employee privacy and safeguarding data security.”</p><h2 id="not-tsmc-39-s-first-rodeo">Not TSMC's first rodeo</h2><p>TSMC, or Taiwan Semiconductor Manufacturing Company to give it its full name, is the world’s largest contract chipmaker, building some of the most advanced semiconductors on the planet to power smartphones and computers, data centers and AI systems, cars and smart home devices, with the likes of Apple, Nvidia and AMD among its clients. </p><p>Its importance to the global economy is enormous, as much of the world's advanced semiconductor supply depends on its manufacturing capabilities. At the end of 2025, it reportedly employed more than 90,000 people and generated around $122.4 billion in revenue (up by more than a third compared to the year before). It also reported approximately $55.2 billion in net profit, making it one of the most financially successful companies in the semiconductor industry.</p><p>This is not TSMC’s first data breach - it has had multiple incidents over the years. In 2018, a malware infection spread across several manufacturing facilities, disrupting production and causing an estimated $85 million losses. </p><p>Three years ago, the LockBit ransomware gang claimed to have stolen data from the company and demanded $70 million, but as it turned out, the data originated from a supplier called Kinmax Technology. </p><p><em>Via </em><a href="https://www.bleepingcomputer.com/news/security/tsmc-employee-data-breach-linked-to-cloud-provider/" target="_blank"><em>BleepingComputer</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/tsmc-confirms-cyberattack-and-data-breach-chip-giant-says-personal-employee-data-has-been-leaked-online</link>
                                                                            <description>
                            <![CDATA[ A threat actor called SmithBianl offered a sample online as they threatened to release the whole batch. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">78DNLYeDt2xdM5EXNxoUnT</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/eeC4nkApykrupZM24rGDnR-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 17:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/eeC4nkApykrupZM24rGDnR-1920-80.jpg">
                                                            <media:credit><![CDATA[Sundry Photography / Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[TSMC]]></media:description>                                                            <media:text><![CDATA[TSMC]]></media:text>
                                <media:title type="plain"><![CDATA[TSMC]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/eeC4nkApykrupZM24rGDnR-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>TSMC confirmed employee data was exposed through a third-party cloud provider rather than its own systems</strong></li><li><strong>Stolen records reportedly include employee personal and banking information, with around 46,000 records offered for sale</strong></li><li><strong>Company is notifying affected personnel and investigating the breach alongside authorities and security experts</strong></li></ul><p>TSMC, the world’s largest semiconductor manufacturer, has confirmed suffering a cyberattack and losing sensitive files after a threat actor offered a large database for sale on the dark web.</p><p>This comes after a threat actor with the alias “SmithBianl” recently posted a new thread on an underground forum offering a text file with 499 rows of data they claim was nabbed from TSMC. The text file, they claim, provides a sample of a much larger dataset they owned, which contains 46,000 employee records and weighs almost 7GB.</p><p>They also said the stolen archive contains TSMC employees’ national ID numbers, addresses, contact details, and banking information, and are asking 3 BTC in exchange (just below $200,000 at current prices). They also said that they would leak the entire archive if a buyer doesn’t step forward by November 15 2026 - a deadline which has now expired. </p><h2 id="third-party-cloud-service-provider-struck">Third-party cloud service provider struck</h2><p>Following the leak, the company’s Senior Vice President and Chief Financial Officer (CFO), Wendell Huang, issued a short statement confirming that the company was not breached and that the stolen files were hosted on a third party:</p><p>“TSMC has found that some of our current and former employee data was made available on an external website without authorization,” the announcement read.</p><p>“Currently, we have found no evidence that TSMC’s systems have been compromised. This information has been verified to match data stored in the system services of a cloud service provider, but is unrelated to the company’s operations or customer-related data.”</p><p>Huang would not say which third-party <a href="https://www.techradar.com/best/best-cloud-storage" target="_blank">cloud provider</a> that was, and the company is currently not responding to media inquiries. Therefore, we also don’t know how that company was breached, and if it hosted - and lost - data belonging to other companies, as well. </p><p>We also don’t know exactly how many people had their data exposed in this incident. 46,000 employee records doesn’t necessarily mean 46,000 people, since victims lost different records such as names and banking information.</p><p>Some of them might be TSMC staff, given the curious choice of words in Huang’s letter. </p><p>“We have been informing impacted personnel and are providing them with necessary support,” he said. “TSMC has engaged with both the cloud service vendor, the relevant investigating authorities, and a third-party cybersecurity agency to jointly ascertain the facts and take necessary actions. We remain firmly committed to protecting employee privacy and safeguarding data security.”</p><h2 id="not-tsmc-39-s-first-rodeo">Not TSMC's first rodeo</h2><p>TSMC, or Taiwan Semiconductor Manufacturing Company to give it its full name, is the world’s largest contract chipmaker, building some of the most advanced semiconductors on the planet to power smartphones and computers, data centers and AI systems, cars and smart home devices, with the likes of Apple, Nvidia and AMD among its clients. </p><p>Its importance to the global economy is enormous, as much of the world's advanced semiconductor supply depends on its manufacturing capabilities. At the end of 2025, it reportedly employed more than 90,000 people and generated around $122.4 billion in revenue (up by more than a third compared to the year before). It also reported approximately $55.2 billion in net profit, making it one of the most financially successful companies in the semiconductor industry.</p><p>This is not TSMC’s first data breach - it has had multiple incidents over the years. In 2018, a malware infection spread across several manufacturing facilities, disrupting production and causing an estimated $85 million losses. </p><p>Three years ago, the LockBit ransomware gang claimed to have stolen data from the company and demanded $70 million, but as it turned out, the data originated from a supplier called Kinmax Technology. </p><p><em>Via </em><a href="https://www.bleepingcomputer.com/news/security/tsmc-employee-data-breach-linked-to-cloud-provider/" target="_blank"><em>BleepingComputer</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Yet another massive cyberattack hits Japan — city of Osaka latest to be hit, as more than a million email addresses exposed ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Osaka City disclosed exposure of 1.08 million email addresses through the wider Arara email platform breach</strong></li><li><strong>Stolen data included addresses used for government notifications, account activity, and administrative services</strong></li><li><strong>Officials warned residents to watch for phishing emails impersonating Osaka City or related online systems</strong></li></ul><p>The Arara email exposure saga continues, as after Asoview, MOS Burger and <a href="https://www.techradar.com/pro/security/sega-newsletter-data-breach-sees-2-2-million-email-addresses-exposed-gaming-giant-says-it-is-investigating-but-user-financial-data-not-affected" target="_blank">Sega</a>, we can now add the City of Osaka to the list of victims as well after it warned of losing more than a million emails and urged citizens to be wary of incoming phishing and other scam emails.</p><p>On November 13, Osaka City also disclosed being affected via a third-party service provider called TKC Corporation, a Japanese IT and software company that provides cloud-based technology services to businesses and government organizations, including accounting systems and public administration platforms. </p><p>The city said whoever received a notification between July 1 and October 31, 2026, has likely had their email address exposed to threat actors:</p><p>“Anyone who received an email from the System (sender address: info-osaka-online@city.osaka.lg.jp) between July 1, 2026 (Wednesday) and October 31, 2026 (Saturday) [is affected],” the machine-translated notice reads.</p><h2 id="tkc-corporation-struck">TKC Corporation struck</h2><p>For Osaka City, TKC provided and operated the Osaka City Administrative Online System, a digital platform through which citizens can register accounts, log in, submit applications, and receive administrative notifications.</p><p>“The leaked information includes the email addresses to which the System sent messages in connection with user registration, changes to email addresses, password changes, logins, applications and other administrative procedures, changes in application processing status made by Osaka City, notifications and other announcements from the System.”</p><p>It is worth mentioning that these are the topics of the emails the city sent to affected individuals - passwords and logins were not exposed themselves. </p><p>“Users of the System are asked to check their email histories to determine whether they received messages from the above sender address between July 1 and October 31, 2026,” Osaka concluded.</p><p>According to the notice, exactly 1,078,713 user email addresses were exposed. The city now urges those citizens to be wary of incoming emails: </p><p>“There is a possibility that suspicious emails related to this incident may be sent,” it said. “Even if an email claims to come from Osaka City or the System, do not open links or attachments in unexpected or unsolicited emails; enter your user ID, password, PIN, credit card details or other personal information, or respond to emails requesting payment.”</p><h2 id="arara-affected-again">Arara affected again</h2><p>The attack is the latest concerning Arara - a technology firm providing digital marketing and customer engagement services. One of its offerings is Arara Message, an <a href="https://www.techradar.com/news/best-email-provider" target="_blank">email distribution platform</a> that businesses can use to send newsletters and other communications to millions of people. Through it, Arara gets to manage contact information on behalf of those organizations - emails, names, phone numbers, and more. </p><p>In early November 2026, Arara reported its email distribution infrastructure was breached. Apparently, a combination of a web server vulnerability and “other system weaknesses” allowed unidentified threat actors to access 97 servers that stored data of up to 380 corporate customers. Obviously, Arara juggled a lot of clients, many of which now have to deal with email leakage.</p><p>Before Osaka City, multiple other companies confirmed being affected: Asoview reported approximately 15.82 million records, while electronics retailer Bic Camera reported 15.2 million. Fast-food chain MOS Burger disclosed the exposure of some 7.5 million email addresses, while cinema operator Shochiku Multiplex Theaters 3.3 million. Sega and its subsidiary Sega Fave disclosed approximately 2.2 million affected email addresses. </p><p>In total, the unidentified actors walked away with more than 81 million email addresses, together with other personally identifiable data. In a subsequent report, Arara added that the attackers compromised a separate environment containing additional 3.38 million email addresses and more than 400,000 customer records containing PII. </p><p>With 380 corporate customers at risk, it’s safe to assume this is not the last time we hear of the Arara breach. </p><p>Via <a href="https://therecord.media/city-of-osaka-japan-email-data-breach" target="_blank"><em>The Record</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/yet-another-massive-cyberattack-hits-japan-city-of-osaka-latest-to-be-hit-as-more-than-a-million-email-addresses-exposed</link>
                                                                            <description>
                            <![CDATA[ Japanese City of Osaka warns citizens of incoming phishing emails following cyberattack. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">5aYkHLRCWgwhgaK8r4aHjT</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/ThNyuwnA55tfcixfqWcEcA-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 16:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/ThNyuwnA55tfcixfqWcEcA-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Pirate skull cyber attack digital technology flag cyber on on computer CPU in background. Darknet and cybercrime banner cyberattack and espionage concept illustration.]]></media:description>                                                            <media:text><![CDATA[Pirate skull cyber attack digital technology flag cyber on on computer CPU in background. Darknet and cybercrime banner cyberattack and espionage concept illustration.]]></media:text>
                                <media:title type="plain"><![CDATA[Pirate skull cyber attack digital technology flag cyber on on computer CPU in background. Darknet and cybercrime banner cyberattack and espionage concept illustration.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/ThNyuwnA55tfcixfqWcEcA-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Osaka City disclosed exposure of 1.08 million email addresses through the wider Arara email platform breach</strong></li><li><strong>Stolen data included addresses used for government notifications, account activity, and administrative services</strong></li><li><strong>Officials warned residents to watch for phishing emails impersonating Osaka City or related online systems</strong></li></ul><p>The Arara email exposure saga continues, as after Asoview, MOS Burger and <a href="https://www.techradar.com/pro/security/sega-newsletter-data-breach-sees-2-2-million-email-addresses-exposed-gaming-giant-says-it-is-investigating-but-user-financial-data-not-affected" target="_blank">Sega</a>, we can now add the City of Osaka to the list of victims as well after it warned of losing more than a million emails and urged citizens to be wary of incoming phishing and other scam emails.</p><p>On November 13, Osaka City also disclosed being affected via a third-party service provider called TKC Corporation, a Japanese IT and software company that provides cloud-based technology services to businesses and government organizations, including accounting systems and public administration platforms. </p><p>The city said whoever received a notification between July 1 and October 31, 2026, has likely had their email address exposed to threat actors:</p><p>“Anyone who received an email from the System (sender address: info-osaka-online@city.osaka.lg.jp) between July 1, 2026 (Wednesday) and October 31, 2026 (Saturday) [is affected],” the machine-translated notice reads.</p><h2 id="tkc-corporation-struck">TKC Corporation struck</h2><p>For Osaka City, TKC provided and operated the Osaka City Administrative Online System, a digital platform through which citizens can register accounts, log in, submit applications, and receive administrative notifications.</p><p>“The leaked information includes the email addresses to which the System sent messages in connection with user registration, changes to email addresses, password changes, logins, applications and other administrative procedures, changes in application processing status made by Osaka City, notifications and other announcements from the System.”</p><p>It is worth mentioning that these are the topics of the emails the city sent to affected individuals - passwords and logins were not exposed themselves. </p><p>“Users of the System are asked to check their email histories to determine whether they received messages from the above sender address between July 1 and October 31, 2026,” Osaka concluded.</p><p>According to the notice, exactly 1,078,713 user email addresses were exposed. The city now urges those citizens to be wary of incoming emails: </p><p>“There is a possibility that suspicious emails related to this incident may be sent,” it said. “Even if an email claims to come from Osaka City or the System, do not open links or attachments in unexpected or unsolicited emails; enter your user ID, password, PIN, credit card details or other personal information, or respond to emails requesting payment.”</p><h2 id="arara-affected-again">Arara affected again</h2><p>The attack is the latest concerning Arara - a technology firm providing digital marketing and customer engagement services. One of its offerings is Arara Message, an <a href="https://www.techradar.com/news/best-email-provider" target="_blank">email distribution platform</a> that businesses can use to send newsletters and other communications to millions of people. Through it, Arara gets to manage contact information on behalf of those organizations - emails, names, phone numbers, and more. </p><p>In early November 2026, Arara reported its email distribution infrastructure was breached. Apparently, a combination of a web server vulnerability and “other system weaknesses” allowed unidentified threat actors to access 97 servers that stored data of up to 380 corporate customers. Obviously, Arara juggled a lot of clients, many of which now have to deal with email leakage.</p><p>Before Osaka City, multiple other companies confirmed being affected: Asoview reported approximately 15.82 million records, while electronics retailer Bic Camera reported 15.2 million. Fast-food chain MOS Burger disclosed the exposure of some 7.5 million email addresses, while cinema operator Shochiku Multiplex Theaters 3.3 million. Sega and its subsidiary Sega Fave disclosed approximately 2.2 million affected email addresses. </p><p>In total, the unidentified actors walked away with more than 81 million email addresses, together with other personally identifiable data. In a subsequent report, Arara added that the attackers compromised a separate environment containing additional 3.38 million email addresses and more than 400,000 customer records containing PII. </p><p>With 380 corporate customers at risk, it’s safe to assume this is not the last time we hear of the Arara breach. </p><p>Via <a href="https://therecord.media/city-of-osaka-japan-email-data-breach" target="_blank"><em>The Record</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Making a typo when trying to access AI could land you on a malware site ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Attackers registered hundreds of typo-squatted domains impersonating popular AI brands such as ChatGPT, Claude, and Perplexity</strong></li><li><strong>Victims who mistyped URLs were redirected to malware, malicious browser extensions, or ClickFix-style attacks</strong></li><li><strong>Researchers warn the campaign remains active, making bookmarks and careful URL verification essential defenses</strong></li></ul><p>Be careful when typing your favorite AI tool’s website into the address bar - one mistake and you might end up with <a href="https://www.techradar.com/best/best-malware-removal" target="_blank">malware</a> on your computer. </p><p>This is the warning given out by security researchers <a href="https://www.island.io/blog/typoprompting-from-ai-typos-to-malware" target="_blank" rel="nofollow">Island</a>, who recently uncovered a large-scale domain squatting campaign spoofing popular AI brands.</p><p>Users who want to open ChatGPT or Claude might misspell the name and end up on sites like chtgpt.com or cluade.ai. In a perfect world, these websites would return a 404 error (page not found) or would at least automatically redirect the visitor to the legitimate website. But we don’t live in a perfect world, but rather in one filled with malicious actors looking to benefit from even the tiniest of mistakes.</p><h2 id="benefitting-from-mistakes">Benefitting from mistakes</h2><p>In its report, Island said it found 186 typo-squatted domains impersonating 27 popular AI brands, among which are your usual household names - ChatGPT, Claude, Perplexity, Midjourney, and Cursor. </p><p>What threat actors did, Island claims, is purchase 186 of these slightly erroneous domains and create websites which, depending on the visitor’s location, operating system and browser, would redirect them to different destinations. On these destinations, one of three things could happen: the victims would be shown a fake Cloudflare verification page, a fake security-check page, or instructions to download various browser extensions. </p><p>All three scenarios are malicious. The first one is a typical click-fix attack that instructs Windows users to copy and execute malicious PowerShell commands. These downloaded and executed additional scripts from attacker-controlled infrastructure but unfortunately, Island does not say what the final payload is or what it does. </p><p>The fake security-check pages offered a seemingly legitimate, Microsoft Store-signed application called Net Guard, which is actually able to hijack the victim’s browser. This one is particularly worrisome, since it was signed and distributed through a package matching the legitimate Microsoft Store listing, and could change default search engines, modify browser profiles, install extensions and execute JavaScript remotely. </p><p>The third attack instructed victims to download malicious extensions that hijacked their default search engine, which allows the attackers to generate profits through monetized searches, or by selling harvested search engine queries. </p><h2 id="are-the-domains-still-active">Are the domains still active?</h2><p>Island says ChatGPT was the most popular company to spoof, with 25 identified lookalikes. It is followed by Claude (18 typosquatted domains), DeepSeek (16), and Runway (12). Perplexity, Lovable and Character.ai all have 11 each, it added. </p><p>It is worth mentioning that this was likely not done by a single actor, but rather numerous ones all using the same technique to target their victims. Also, Island stressed that the campaign likely extended beyond AI services - after examining shared infrastructure, the researchers connected the operation to 211 entry domains, including lookalikes of banks, insurers, telecom companies and enterprise software providers. </p><p>Security researchers usually publish these reports after helping minimize the risk which, in this case, would be seeing the domains seized and the malicious websites brought offline. </p><p>However, Island’s report does not discuss any takedowns, so it is safe to assume at least some of these domains are still active today. As a rule of thumb, make sure to carefully review the domain typed into the address bar before navigating anywhere. Alternatively, bookmarking your favorite tools is a good option, as well. </p><p>Malicious browser extensions did not get the same treatment, though. Of the 30 identified ones, 13 have already been removed from the Chrome Web Store, while 17 remained listed as of November 3, 2026. If not already, we can expect them to be removed in the coming days and weeks. </p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/making-a-typo-when-trying-to-access-ai-could-land-you-on-a-malware-site</link>
                                                                            <description>
                            <![CDATA[ Popular AI brands abused in a new typosquatting campaign, researchers warn. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">SGo6PHnRD3NGsVFTmAUfQg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/8iN8JcsjaeEppccYgURuY9-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 15:18:45 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/8iN8JcsjaeEppccYgURuY9-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[The ChatGPT and Claude mobile apps on a phone screen next to a digitized image of a robot]]></media:description>                                                            <media:text><![CDATA[The ChatGPT and Claude mobile apps on a phone screen next to a digitized image of a robot]]></media:text>
                                <media:title type="plain"><![CDATA[The ChatGPT and Claude mobile apps on a phone screen next to a digitized image of a robot]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/8iN8JcsjaeEppccYgURuY9-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Attackers registered hundreds of typo-squatted domains impersonating popular AI brands such as ChatGPT, Claude, and Perplexity</strong></li><li><strong>Victims who mistyped URLs were redirected to malware, malicious browser extensions, or ClickFix-style attacks</strong></li><li><strong>Researchers warn the campaign remains active, making bookmarks and careful URL verification essential defenses</strong></li></ul><p>Be careful when typing your favorite AI tool’s website into the address bar - one mistake and you might end up with <a href="https://www.techradar.com/best/best-malware-removal" target="_blank">malware</a> on your computer. </p><p>This is the warning given out by security researchers <a href="https://www.island.io/blog/typoprompting-from-ai-typos-to-malware" target="_blank" rel="nofollow">Island</a>, who recently uncovered a large-scale domain squatting campaign spoofing popular AI brands.</p><p>Users who want to open ChatGPT or Claude might misspell the name and end up on sites like chtgpt.com or cluade.ai. In a perfect world, these websites would return a 404 error (page not found) or would at least automatically redirect the visitor to the legitimate website. But we don’t live in a perfect world, but rather in one filled with malicious actors looking to benefit from even the tiniest of mistakes.</p><h2 id="benefitting-from-mistakes">Benefitting from mistakes</h2><p>In its report, Island said it found 186 typo-squatted domains impersonating 27 popular AI brands, among which are your usual household names - ChatGPT, Claude, Perplexity, Midjourney, and Cursor. </p><p>What threat actors did, Island claims, is purchase 186 of these slightly erroneous domains and create websites which, depending on the visitor’s location, operating system and browser, would redirect them to different destinations. On these destinations, one of three things could happen: the victims would be shown a fake Cloudflare verification page, a fake security-check page, or instructions to download various browser extensions. </p><p>All three scenarios are malicious. The first one is a typical click-fix attack that instructs Windows users to copy and execute malicious PowerShell commands. These downloaded and executed additional scripts from attacker-controlled infrastructure but unfortunately, Island does not say what the final payload is or what it does. </p><p>The fake security-check pages offered a seemingly legitimate, Microsoft Store-signed application called Net Guard, which is actually able to hijack the victim’s browser. This one is particularly worrisome, since it was signed and distributed through a package matching the legitimate Microsoft Store listing, and could change default search engines, modify browser profiles, install extensions and execute JavaScript remotely. </p><p>The third attack instructed victims to download malicious extensions that hijacked their default search engine, which allows the attackers to generate profits through monetized searches, or by selling harvested search engine queries. </p><h2 id="are-the-domains-still-active">Are the domains still active?</h2><p>Island says ChatGPT was the most popular company to spoof, with 25 identified lookalikes. It is followed by Claude (18 typosquatted domains), DeepSeek (16), and Runway (12). Perplexity, Lovable and Character.ai all have 11 each, it added. </p><p>It is worth mentioning that this was likely not done by a single actor, but rather numerous ones all using the same technique to target their victims. Also, Island stressed that the campaign likely extended beyond AI services - after examining shared infrastructure, the researchers connected the operation to 211 entry domains, including lookalikes of banks, insurers, telecom companies and enterprise software providers. </p><p>Security researchers usually publish these reports after helping minimize the risk which, in this case, would be seeing the domains seized and the malicious websites brought offline. </p><p>However, Island’s report does not discuss any takedowns, so it is safe to assume at least some of these domains are still active today. As a rule of thumb, make sure to carefully review the domain typed into the address bar before navigating anywhere. Alternatively, bookmarking your favorite tools is a good option, as well. </p><p>Malicious browser extensions did not get the same treatment, though. Of the 30 identified ones, 13 have already been removed from the Chrome Web Store, while 17 remained listed as of November 3, 2026. If not already, we can expect them to be removed in the coming days and weeks. </p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ AI is reshaping the creator economy: trust will decide who wins it ]]></title>
                                                                                                <dc:content><![CDATA[ <p>There’s always been a balancing act between creators and creator platforms over profit, value, and ownership. Now, <a href="https://www.techradar.com/best/best-ai-tools">AI tools</a> are upsetting that balance and reshaping commercial dynamics, while raising new concerns about transparency and control.</p><p>As creators compete for engagement against low-effort AI “slop,” they must also navigate opaque monetization policies and inconsistent enforcement. These pressures are rapidly increasing the friction between creators and the platforms they rely on. </p><p>At the same time, as more platforms and revenue streams emerge, creators have greater choice about where to build an audience.</p><p>As the terms creators operate under keep changing, creators are pushing for more control. They want their voices to be heard, their work protected, and the value they generate recognized. </p><p>For platforms, this raises the stakes for attracting and retaining top creators. Trust is becoming a competitive differentiator, and the platforms that recognize this are protecting their creators and paying them reliably in the age of AI.</p><h2 id="the-creator-economy-s-ongoing-transparency-issue">The creator economy’s ongoing transparency issue</h2><p>Today, high-volume, low-effort AI-generated content has become an inescapable part of the online experience. But for creators, lax moderation of that content can directly impact earnings by creating an uneven playing field and increasing competition for engagement and views.</p><p>Not only must creators compete with AI-generated content, but concerns are also mounting that their work is being used to train the models producing it, without their consent or compensation. Trust is eroding as a result, and inaction risks alienating the very people whose work underpins the success of digital platforms.</p><p>These challenges are becoming increasingly evident. Recently, a US court approved a $1.5 billion settlement relating to Anthropic’s use of pirated books to train its AI models. While the case represents a clarification for rights holders, smaller creators and artists often lack the resources required to pursue litigation in the first place.</p><p>Concerns have also been raised about YouTube reportedly using AI to alter creators’ <a href="https://www.techradar.com/best/best-video-editing-software">videos</a> without their consent. Even where there is no evidence of financial loss, the incident highlights a potential broader imbalance in the relationship if platforms can make unilateral changes to creators’ work without meaningful consultation.</p><p>Creators rely on platforms to distribute, monetize, and govern their work fairly. But platforms need creators too. Their <a href="https://www.techradar.com/best/best-business-plan-software">business</a> models depend on compelling content that attracts and retains audiences. It’s a mutual dependency that makes listening to and responding to creators’ concerns a strategic imperative.</p><h2 id="creator-trust-is-intrinsically-linked-with-income-reliability">Creator trust is intrinsically linked with income reliability</h2><p>Questions about trust extend beyond how content is used to how creators earn a living. In just a few decades, the creator economy has evolved from a relatively informal ecosystem into a multibillion-dollar global industry. The decisions made by the platforms shaping this industry now directly affect how millions of creators build their careers and earn income.</p><p>Even before AI, changes to recommendation algorithms, monetization models, and advertising guidelines meant creators had to become accustomed to their earnings changing overnight. These realities have encouraged them to move away from relying on individual platforms for predictable income and monetizing their content and brands across multiple platforms and revenue streams instead.</p><p>Whether they earn through commission-based selling on <a href="https://www.techradar.com/news/the-best-ecommerce-platform">ecommerce platforms</a>, livestreaming, brand sponsorships, or content views, creators’ incomes are becoming increasingly fragmented. While this diversification reduces their dependence on any single platform, it also creates greater complexity for the businesses responsible for managing and paying them.</p><p>Across the industry, poor payout experiences are already damaging creator relationships. Tipalti’s recent global <a href="https://www.techradar.com/best/best-payment-gateways">payments</a> research revealed that 26% of businesses lost contributors because of payment-related issues in the past 12 months, with creators accounting for 36% of those losses. Payment friction can directly affect a platform’s ability to retain the people generating its content.</p><p>Failing to address these pain points creates a commercial risk, particularly when creators have more options than ever to build and monetize their audiences. If a platform is frustrating to work with or fails to provide reliable payouts and transparent breakdowns of earnings, creators can move to a competitor that offers a better <a href="https://www.techradar.com/best/cx-tools">customer experience</a>.</p><h2 id="the-foundations-of-creator-trust">The foundations of creator trust</h2><p>To keep creators engaged and earn their trust, platforms must address two priorities: protecting creators' work and paying them reliably and transparently.</p><p>First, creators need confidence that their interests are protected and that they remain in control of their work. Platforms must acknowledge the issues posed by AI, introduce safeguards for creators’ content, and ensure that any resulting benefits are shared fairly.</p><p>Second, platforms must offer a transparent and reliable payout experience. At the end of each payment cycle, creators should receive a clear breakdown of how their payout was calculated, including the factors responsible for any monthly variations. They should feel confident that the economic value they create is recognized and that they will be paid accurately and on time.</p><p>Both priorities carry more weight as creator economies grow worldwide, presenting new opportunities for global partnerships. New markets often bring inconsistent standards for content protection and payments, and incompatible financial infrastructure can bottleneck growth. </p><p>To compete for creators in these regions, platforms need protections that hold up regardless of where a creator is based, and payout infrastructure capable of supporting an expanding range of currencies, payment methods, and regional requirements.</p><h2 id="the-next-stage-of-the-creator-economy">The next stage of the creator economy</h2><p>AI is changing what creators make, how platforms distribute it, and how value gets created. But the fundamental relationship hasn’t changed: creators need platforms to reach audiences, and platforms need creators to keep those audiences coming back.</p><p>As that relationship becomes more complex, trust will matter more, not less. Creators need to know their work is protected, their earnings are transparent, and they’ll be paid accurately and on time, no matter where they are. These aren't minor priorities. They show whether a platform actually values its creators, or just profits from them.</p><p>In the age of AI, trust won’t just shape the creator economy. It will shape who succeeds in it.</p><p><em></em><a href="https://www.techradar.com/news/best-mobile-payment-app"><em>We've featured the best mobile payment apps</em></a><em>.</em></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/ai-is-reshaping-the-creator-economy-trust-will-decide-who-wins-it</link>
                                                                            <description>
                            <![CDATA[ Why protecting creators and providing reliable payments are becoming critical to platform success. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">5GPKaTn3bJFJRFYH3rFcuM</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/4d3FzfBhbeGTkD9mnMpEdM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 15:13:41 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Rob Israch ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/4d3FzfBhbeGTkD9mnMpEdM-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A close up of a person&#039;s eyes and face. They are wearing glasses and in one eye there&#039;s. a reflection of a digital brain]]></media:description>                                                            <media:text><![CDATA[A close up of a person&#039;s eyes and face. They are wearing glasses and in one eye there&#039;s. a reflection of a digital brain]]></media:text>
                                <media:title type="plain"><![CDATA[A close up of a person&#039;s eyes and face. They are wearing glasses and in one eye there&#039;s. a reflection of a digital brain]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/4d3FzfBhbeGTkD9mnMpEdM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>There’s always been a balancing act between creators and creator platforms over profit, value, and ownership. Now, <a href="https://www.techradar.com/best/best-ai-tools">AI tools</a> are upsetting that balance and reshaping commercial dynamics, while raising new concerns about transparency and control.</p><p>As creators compete for engagement against low-effort AI “slop,” they must also navigate opaque monetization policies and inconsistent enforcement. These pressures are rapidly increasing the friction between creators and the platforms they rely on. </p><p>At the same time, as more platforms and revenue streams emerge, creators have greater choice about where to build an audience.</p><p>As the terms creators operate under keep changing, creators are pushing for more control. They want their voices to be heard, their work protected, and the value they generate recognized. </p><p>For platforms, this raises the stakes for attracting and retaining top creators. Trust is becoming a competitive differentiator, and the platforms that recognize this are protecting their creators and paying them reliably in the age of AI.</p><h2 id="the-creator-economy-s-ongoing-transparency-issue">The creator economy’s ongoing transparency issue</h2><p>Today, high-volume, low-effort AI-generated content has become an inescapable part of the online experience. But for creators, lax moderation of that content can directly impact earnings by creating an uneven playing field and increasing competition for engagement and views.</p><p>Not only must creators compete with AI-generated content, but concerns are also mounting that their work is being used to train the models producing it, without their consent or compensation. Trust is eroding as a result, and inaction risks alienating the very people whose work underpins the success of digital platforms.</p><p>These challenges are becoming increasingly evident. Recently, a US court approved a $1.5 billion settlement relating to Anthropic’s use of pirated books to train its AI models. While the case represents a clarification for rights holders, smaller creators and artists often lack the resources required to pursue litigation in the first place.</p><p>Concerns have also been raised about YouTube reportedly using AI to alter creators’ <a href="https://www.techradar.com/best/best-video-editing-software">videos</a> without their consent. Even where there is no evidence of financial loss, the incident highlights a potential broader imbalance in the relationship if platforms can make unilateral changes to creators’ work without meaningful consultation.</p><p>Creators rely on platforms to distribute, monetize, and govern their work fairly. But platforms need creators too. Their <a href="https://www.techradar.com/best/best-business-plan-software">business</a> models depend on compelling content that attracts and retains audiences. It’s a mutual dependency that makes listening to and responding to creators’ concerns a strategic imperative.</p><h2 id="creator-trust-is-intrinsically-linked-with-income-reliability">Creator trust is intrinsically linked with income reliability</h2><p>Questions about trust extend beyond how content is used to how creators earn a living. In just a few decades, the creator economy has evolved from a relatively informal ecosystem into a multibillion-dollar global industry. The decisions made by the platforms shaping this industry now directly affect how millions of creators build their careers and earn income.</p><p>Even before AI, changes to recommendation algorithms, monetization models, and advertising guidelines meant creators had to become accustomed to their earnings changing overnight. These realities have encouraged them to move away from relying on individual platforms for predictable income and monetizing their content and brands across multiple platforms and revenue streams instead.</p><p>Whether they earn through commission-based selling on <a href="https://www.techradar.com/news/the-best-ecommerce-platform">ecommerce platforms</a>, livestreaming, brand sponsorships, or content views, creators’ incomes are becoming increasingly fragmented. While this diversification reduces their dependence on any single platform, it also creates greater complexity for the businesses responsible for managing and paying them.</p><p>Across the industry, poor payout experiences are already damaging creator relationships. Tipalti’s recent global <a href="https://www.techradar.com/best/best-payment-gateways">payments</a> research revealed that 26% of businesses lost contributors because of payment-related issues in the past 12 months, with creators accounting for 36% of those losses. Payment friction can directly affect a platform’s ability to retain the people generating its content.</p><p>Failing to address these pain points creates a commercial risk, particularly when creators have more options than ever to build and monetize their audiences. If a platform is frustrating to work with or fails to provide reliable payouts and transparent breakdowns of earnings, creators can move to a competitor that offers a better <a href="https://www.techradar.com/best/cx-tools">customer experience</a>.</p><h2 id="the-foundations-of-creator-trust">The foundations of creator trust</h2><p>To keep creators engaged and earn their trust, platforms must address two priorities: protecting creators' work and paying them reliably and transparently.</p><p>First, creators need confidence that their interests are protected and that they remain in control of their work. Platforms must acknowledge the issues posed by AI, introduce safeguards for creators’ content, and ensure that any resulting benefits are shared fairly.</p><p>Second, platforms must offer a transparent and reliable payout experience. At the end of each payment cycle, creators should receive a clear breakdown of how their payout was calculated, including the factors responsible for any monthly variations. They should feel confident that the economic value they create is recognized and that they will be paid accurately and on time.</p><p>Both priorities carry more weight as creator economies grow worldwide, presenting new opportunities for global partnerships. New markets often bring inconsistent standards for content protection and payments, and incompatible financial infrastructure can bottleneck growth. </p><p>To compete for creators in these regions, platforms need protections that hold up regardless of where a creator is based, and payout infrastructure capable of supporting an expanding range of currencies, payment methods, and regional requirements.</p><h2 id="the-next-stage-of-the-creator-economy">The next stage of the creator economy</h2><p>AI is changing what creators make, how platforms distribute it, and how value gets created. But the fundamental relationship hasn’t changed: creators need platforms to reach audiences, and platforms need creators to keep those audiences coming back.</p><p>As that relationship becomes more complex, trust will matter more, not less. Creators need to know their work is protected, their earnings are transparent, and they’ll be paid accurately and on time, no matter where they are. These aren't minor priorities. They show whether a platform actually values its creators, or just profits from them.</p><p>In the age of AI, trust won’t just shape the creator economy. It will shape who succeeds in it.</p><p><em></em><a href="https://www.techradar.com/news/best-mobile-payment-app"><em>We've featured the best mobile payment apps</em></a><em>.</em></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Microsoft is hiking cloud prices in India — but dropping costs in China and other markets ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Microsoft raising cloud prices for customers in India, New Zealand by 15%, 10%</strong></li><li><strong>Currency fluctuations have afforded three countries to get cheaper cloud prices</strong></li><li><strong>Brazil changes also on the cards, Japan announcement expected later</strong></li></ul><p>Microsoft has confirmed the details of its latest currency adjustments in a bid to keep local pricing more closely aligned with its global pricing levels and to account for exchange rate fluctuations, and it's bad news for India which is set to see cloud pricing increases of 15.7%.</p><p>New Zealand will also see a 10% rise, while the two nations will also see rises of up to 20% and 10% across on-prem pricing.</p><p>But not all countries are seeing price hikes – China (-8.5%), Norway (-5.9%) and Australia (-4%) will each see a decrease in cloud pricing in line with currency exchange rates. These three countries see no change to on-prem pricing.</p><h2 id="microsoft-announces-its-latest-exchange-rate-linked-pricing-adjustments">Microsoft announces its latest exchange rate-linked pricing adjustments</h2><p>It's worth noting that individual products may see slightly different adjustments based on rounding rules and also how the prices are calculated, but the above percentages offer a preview into the sort of changes customers in these five nations can expect to see in the new year.</p><p>The company confirmed in a <a href="https://news.microsoft.com/source/2026/11/10/annual-currency-adjustments-to-microsoft-commercial-cloud-pricing/" target="_blank" rel="nofollow">blog post</a> that the changes will come into effect from January 1, 2027, as part of a new scheme it introduced in July 2026 whereby the hyperscaler would adjust prices each January for greater predictability, rather than at random points throughout the year.</p><p>Microsoft is also looking to adjust its Brazil pricing following the implementation of the nation's tax reform, while Japan will remain on its previous annual April cadence (with an announcement on track for February 2027).</p><p>It's argues that the revisions will help maintain competitive and regionally consistent prices for its cloud services, but some customers worry that currency fluctuations can make identical services more expensive or cheaper depending on where customers are billed.</p><p>Crucially though, these adjustments are purely a response to global currency fluctuations, and it's not a sign that Microsoft is raising its prices across the board.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/microsoft-is-hiking-cloud-prices-in-india-but-dropping-costs-in-china-and-other-markets</link>
                                                                            <description>
                            <![CDATA[ India and New Zealand customers set to pay at least 10% more, but three countries are seeing cost reductions. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">s7FqeA5tPxQNZA5sL8EJpX</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/usYAYYVM6TwaqQfKykCGSV-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 14:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Cloud Computing]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/usYAYYVM6TwaqQfKykCGSV-1920-80.jpg">
                                                            <media:credit><![CDATA[Microsoft]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Microsoft 365 Business app logos]]></media:description>                                                            <media:text><![CDATA[Microsoft 365 Business app logos]]></media:text>
                                <media:title type="plain"><![CDATA[Microsoft 365 Business app logos]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/usYAYYVM6TwaqQfKykCGSV-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Microsoft raising cloud prices for customers in India, New Zealand by 15%, 10%</strong></li><li><strong>Currency fluctuations have afforded three countries to get cheaper cloud prices</strong></li><li><strong>Brazil changes also on the cards, Japan announcement expected later</strong></li></ul><p>Microsoft has confirmed the details of its latest currency adjustments in a bid to keep local pricing more closely aligned with its global pricing levels and to account for exchange rate fluctuations, and it's bad news for India which is set to see cloud pricing increases of 15.7%.</p><p>New Zealand will also see a 10% rise, while the two nations will also see rises of up to 20% and 10% across on-prem pricing.</p><p>But not all countries are seeing price hikes – China (-8.5%), Norway (-5.9%) and Australia (-4%) will each see a decrease in cloud pricing in line with currency exchange rates. These three countries see no change to on-prem pricing.</p><h2 id="microsoft-announces-its-latest-exchange-rate-linked-pricing-adjustments">Microsoft announces its latest exchange rate-linked pricing adjustments</h2><p>It's worth noting that individual products may see slightly different adjustments based on rounding rules and also how the prices are calculated, but the above percentages offer a preview into the sort of changes customers in these five nations can expect to see in the new year.</p><p>The company confirmed in a <a href="https://news.microsoft.com/source/2026/11/10/annual-currency-adjustments-to-microsoft-commercial-cloud-pricing/" target="_blank" rel="nofollow">blog post</a> that the changes will come into effect from January 1, 2027, as part of a new scheme it introduced in July 2026 whereby the hyperscaler would adjust prices each January for greater predictability, rather than at random points throughout the year.</p><p>Microsoft is also looking to adjust its Brazil pricing following the implementation of the nation's tax reform, while Japan will remain on its previous annual April cadence (with an announcement on track for February 2027).</p><p>It's argues that the revisions will help maintain competitive and regionally consistent prices for its cloud services, but some customers worry that currency fluctuations can make identical services more expensive or cheaper depending on where customers are billed.</p><p>Crucially though, these adjustments are purely a response to global currency fluctuations, and it's not a sign that Microsoft is raising its prices across the board.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ 'Every organization has to invest in getting this right': Proofpoint CEO warns the agentic AI age will mean a whole new generation of cyber threats ]]></title>
                                                                                                <dc:content><![CDATA[ <p>The CEO of Proofpoint has urged companies of all sizes to be on guard against new and advanced cyber threats built and enabled by the rise of AI.</p><p>Speaking at the company’s recent Proofpoint Protect Tour London 2026 event, Sumit Dhawan warned that customers should, “think of AI agents as insiders in your environment” as the number of AI-assisted attacks continues to rise.</p><p>The company recently released research claiming <a href="https://www.techradar.com/pro/security/european-cyber-threats-are-growing-much-faster-than-other-regions-as-new-report-warns-ai-enables-cybercriminals-to-operate-at-a-scale-weve-never-seen-before" target="_blank">AI enables cybercriminals to operate at a scale 'never seen before'</a> - so how does it suggest companies stay safe?</p><h2 id="defend-and-access-with-intent">Defend and Access with Intent</h2><p>In his keynote, Dhawan declared that “the concept of time in cybersecurity keeps shrinking”, giving the example of a recent cyber heist which took place in just 27 seconds.</p><p>“An agentic toolkit is now able to identify vulnerabilities, issue attacks, and create multiple social engineering campaigns in an automated fashion,” he added, “in this world, we believe that making sure that your human perimeter is protected, because reachability becomes one of the biggest things - and the defence needs to be done with proper understanding of intent.”</p><p>Dhawan outlined Proofpoint’s strategy of Defend with Intent, and Access with Intent, the former looking to understand where attacks originate and how to build resiliency and defense, especially with AI giving attackers a whole new arsenal of tools. </p><p>The latter looks to ensure humans and AI can continue to work effectively together, as agents don’t access data or apps the same way humans do, necessitating a whole new approach to make sure company systems are safe.</p><h2 id="quot-ai-journey-quot">"AI journey"</h2><p>In a later media Q&amp;A session, Dhawan and Daniel Blackford, head of threat research at Proofpoint, expanded on the role that AI is playing in these new threats, noting how quickly the threat landscape has changed.</p><p>“I think things have changed,” Blackford says, “threat actors have gone on a similar AI journey that a lot of us have,” although he notes that there is still an adoption and learning curve, as attackers and defenders alike try to discover the best way to utilize AI.</p><p>“The problem I see is not whether it's hype or not - but what you do about it,” Dhawan adds, noting how the need for constant and rapid reaction to threats is quickly going to become fatiguing for workers - so the crucial consideration is what businesses do about it.</p><p>“It's somewhat of a thankless job, you're always chasing a major threat,” he notes, “ (but) every organization has to invest in getting the right configuration of access and data.”</p><p>“Defenders have had a window where we've been ahead of where threat actors were with using AI - but that window is shrinking,” adds Blackford.</p><p>This may be particularly key for small businesses - and I ask Dhawan and Blackford what advice they have for SMBs looking to stay ahead of the latest threats. </p><p>“For small businesses, in the past, cybersecurity for them was a little bit of basic hygiene - do I have my antivirus, my firewall, some kind of email spam filtering that also catches some of these threats - now, it’s a real issue,” Dhawan says, pointing to Proofpoint’s recent acquisition of Hornetsecurity as its bid to offer thorough protection for SMBs.</p><p>“It’s quite important for small and medium businesses to take (cybersecurity) more seriously now than they ever have,” Blackford adds, noting that these firms could suffer immensely from a lack of reputation due to suffering an attack.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1080px;"><p class="vanilla-image-block" style="padding-top:55.56%;"><img id="uTLwBhC26YCauAq8Swffd8" name="cybersecurity-2.jpg" alt="A computer being guarded by cybersecurity." src="https://cdn.mos.cms.futurecdn.net/uTLwBhC26YCauAq8Swffd8-1920-80.jpg" mos="" align="middle" fullscreen="" width="1080" height="600" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: iStock)</span></figcaption></figure><p>Ultimately, Proofpoint sees its mission as attempting to protect its customers from the biggest number of threats it can - with its recent acquisition of Varonis demonstrating its intent to offer AI-related protection like never before.</p><p>“Protection is not meaningful if it doesn't work with everything you have in your environment,” Dhawan noted, pointing out that as the average business can now see millions of AI prompts and actions in a typical year - a mammoth task which simply cannot be manually audited and investigated by humans alone.</p><p>“We're doing our part of protecting the behavior, and ensuring your data is connected when people and AI work in this agentic workspace,” Dhawan declared in his speech, “our goal will always be that we integrate (and) by having focus, we are able to do what we do best.”</p><p>“This is the power of AI and the community working together to defeat threat actors.”</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/every-organization-has-to-invest-in-getting-this-right-proofpoint-ceo-warns-the-agentic-ai-age-will-mean-a-whole-new-generation-of-cyber-threats</link>
                                                                            <description>
                            <![CDATA[ As AI-enabled cyber threats continue to emerge, Proofpoint tells us its plan to keep customers protected. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">56yTYnDRZnFJTuP4b4FUyg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/DeMQZ2PBM9wVwRGjayqtMm-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 13:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                                                                                    <dc:creator><![CDATA[ Mike Moore ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/vinm2oPWMvB8yMg7qLhtxg-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Mike Moore is Deputy Editor at TechRadar Pro. He has worked as a B2B and B2C technology journalist for over a decade, including at one of the UK&#039;s leading national newspapers and fellow Future title ITProPortal, covering everything from cybersecurity to phone reviews to VR at the Winter Olympics.&lt;/p&gt;&lt;p&gt;&lt;br&gt;&lt;/p&gt;&lt;p&gt;Mike is the main editorial contact for TechRadar Pro, responsible for the news content across the site, as well as managing the contributed content. PRs looking to pitch news stories, bylines/analysis pieces or event invitations should get in contact via the email address mentioned above.&lt;/p&gt;&lt;p&gt;&lt;br&gt;&lt;/p&gt;&lt;p&gt;He has a Masters degree in American Studies from the University of Nottingham, along with a BA in American &amp; English Studies from the same institution. When he&#039;s not keeping track of all the latest enterprise and workplace trends, he can most likely be found watching, following or taking part in some kind of sport.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/DeMQZ2PBM9wVwRGjayqtMm-1920-80.jpg">
                                                            <media:credit><![CDATA[Future/Mike Moore]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Proofpoint Protect Tour London 2026]]></media:description>                                                            <media:text><![CDATA[Proofpoint Protect Tour London 2026]]></media:text>
                                <media:title type="plain"><![CDATA[Proofpoint Protect Tour London 2026]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/DeMQZ2PBM9wVwRGjayqtMm-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>The CEO of Proofpoint has urged companies of all sizes to be on guard against new and advanced cyber threats built and enabled by the rise of AI.</p><p>Speaking at the company’s recent Proofpoint Protect Tour London 2026 event, Sumit Dhawan warned that customers should, “think of AI agents as insiders in your environment” as the number of AI-assisted attacks continues to rise.</p><p>The company recently released research claiming <a href="https://www.techradar.com/pro/security/european-cyber-threats-are-growing-much-faster-than-other-regions-as-new-report-warns-ai-enables-cybercriminals-to-operate-at-a-scale-weve-never-seen-before" target="_blank">AI enables cybercriminals to operate at a scale 'never seen before'</a> - so how does it suggest companies stay safe?</p><h2 id="defend-and-access-with-intent">Defend and Access with Intent</h2><p>In his keynote, Dhawan declared that “the concept of time in cybersecurity keeps shrinking”, giving the example of a recent cyber heist which took place in just 27 seconds.</p><p>“An agentic toolkit is now able to identify vulnerabilities, issue attacks, and create multiple social engineering campaigns in an automated fashion,” he added, “in this world, we believe that making sure that your human perimeter is protected, because reachability becomes one of the biggest things - and the defence needs to be done with proper understanding of intent.”</p><p>Dhawan outlined Proofpoint’s strategy of Defend with Intent, and Access with Intent, the former looking to understand where attacks originate and how to build resiliency and defense, especially with AI giving attackers a whole new arsenal of tools. </p><p>The latter looks to ensure humans and AI can continue to work effectively together, as agents don’t access data or apps the same way humans do, necessitating a whole new approach to make sure company systems are safe.</p><h2 id="quot-ai-journey-quot">"AI journey"</h2><p>In a later media Q&amp;A session, Dhawan and Daniel Blackford, head of threat research at Proofpoint, expanded on the role that AI is playing in these new threats, noting how quickly the threat landscape has changed.</p><p>“I think things have changed,” Blackford says, “threat actors have gone on a similar AI journey that a lot of us have,” although he notes that there is still an adoption and learning curve, as attackers and defenders alike try to discover the best way to utilize AI.</p><p>“The problem I see is not whether it's hype or not - but what you do about it,” Dhawan adds, noting how the need for constant and rapid reaction to threats is quickly going to become fatiguing for workers - so the crucial consideration is what businesses do about it.</p><p>“It's somewhat of a thankless job, you're always chasing a major threat,” he notes, “ (but) every organization has to invest in getting the right configuration of access and data.”</p><p>“Defenders have had a window where we've been ahead of where threat actors were with using AI - but that window is shrinking,” adds Blackford.</p><p>This may be particularly key for small businesses - and I ask Dhawan and Blackford what advice they have for SMBs looking to stay ahead of the latest threats. </p><p>“For small businesses, in the past, cybersecurity for them was a little bit of basic hygiene - do I have my antivirus, my firewall, some kind of email spam filtering that also catches some of these threats - now, it’s a real issue,” Dhawan says, pointing to Proofpoint’s recent acquisition of Hornetsecurity as its bid to offer thorough protection for SMBs.</p><p>“It’s quite important for small and medium businesses to take (cybersecurity) more seriously now than they ever have,” Blackford adds, noting that these firms could suffer immensely from a lack of reputation due to suffering an attack.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:1080px;"><p class="vanilla-image-block" style="padding-top:55.56%;"><img id="uTLwBhC26YCauAq8Swffd8" name="cybersecurity-2.jpg" alt="A computer being guarded by cybersecurity." src="https://cdn.mos.cms.futurecdn.net/uTLwBhC26YCauAq8Swffd8-1920-80.jpg" mos="" align="middle" fullscreen="" width="1080" height="600" attribution="" endorsement="" class="inline" /></p></div></div><figcaption itemprop="caption description" class=" inline-layout"><span class="credit" itemprop="copyrightHolder">(Image credit: iStock)</span></figcaption></figure><p>Ultimately, Proofpoint sees its mission as attempting to protect its customers from the biggest number of threats it can - with its recent acquisition of Varonis demonstrating its intent to offer AI-related protection like never before.</p><p>“Protection is not meaningful if it doesn't work with everything you have in your environment,” Dhawan noted, pointing out that as the average business can now see millions of AI prompts and actions in a typical year - a mammoth task which simply cannot be manually audited and investigated by humans alone.</p><p>“We're doing our part of protecting the behavior, and ensuring your data is connected when people and AI work in this agentic workspace,” Dhawan declared in his speech, “our goal will always be that we integrate (and) by having focus, we are able to do what we do best.”</p><p>“This is the power of AI and the community working together to defeat threat actors.”</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Replacing workers with AI could prove incredibly costly if the technology doesn't work out ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Workers said they would demand more pay and benefits when returning to their employer following an unsuccessful AI trial</strong></li><li><strong>Half of workers would actually prioritize security over titles and perks</strong></li><li><strong>Reputational damage and worker loyalty could also affect employers</strong></li></ul><p>New research has warned companies looking to replace human workers with AI could face major reputational consequences later on down the line if their AI strategies fall apart.</p><p>Companies that are forced to re-hire human workers after laying them off could ultimately find themselves having to pay higher salaries, offer better contractual terms and include stronger employment protections, the report from JobAssist warns.</p><p>But even after offering workers better terms overall, they could have already made reputational and loyalty damages – nearly one-third (29%) of workers say they'd return to their old job, but would immediately start looking for another.</p><h2 id="replacing-humans-with-ai-could-have-long-lasting-consequences">Replacing humans with AI could have long-lasting consequences</h2><p>While a relatively low but not insignificant number of workers would likely continue to look for other opportunities, the number of workers looking for a higher salary (71%) is much higher. Nearly two in three (64%) also said they'd negotiate more aggressively in general when returning to an old job.</p><p>However, workers are still very clearly worried about AI's impacts on their roles, because half (54%) would now prioritize job security over titles and perks, implying they're concerned about experiencing yet another redundancy. Nearly half (44%) also said they'd trust their employer less after AI replaced human workers.</p><p>All in all, the study is proof that employer reputation could suffer beyond the original workforce reduction, with laid off workers and re-hires less likely to trust the company. </p><p>And even after all the costs associated with laying off workers in the first place, and investing so heavily in artificial intelligence, JobAssist warns that workers could end up demanding more money, benefits and security than they had done before.</p><p>"For businesses evaluating the financial benefits of replacing employees with AI, the survey highlights a risk that could emerge if an automation decision has to be reversed," the report summarized.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/their-relationship-with-the-company-could-be-permanently-altered-bosses-beware-replacing-workers-with-ai-could-prove-incredibly-costly-if-the-technology-doesnt-work-out</link>
                                                                            <description>
                            <![CDATA[ Companies who replace human workers with AI could face costly challenges if they're forced to re-hire their old workers. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">rV9XxtQBLvQ4gjAreQ3zgC</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/jYVTL37ZnmoRo5trgcQewG-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 12:05:00 +0000</pubDate>                                                                                                                                <updated>Tue, 17 Nov 2026 05:14:49 +0000</updated>
                                                                                                                                            <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/jYVTL37ZnmoRo5trgcQewG-1920-80.png">
                                                            <media:credit><![CDATA[Unsplash / Alex Kotliarskyi]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Office with many desks and workers at computers]]></media:description>                                                            <media:text><![CDATA[Office with many desks and workers at computers]]></media:text>
                                <media:title type="plain"><![CDATA[Office with many desks and workers at computers]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/jYVTL37ZnmoRo5trgcQewG-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Workers said they would demand more pay and benefits when returning to their employer following an unsuccessful AI trial</strong></li><li><strong>Half of workers would actually prioritize security over titles and perks</strong></li><li><strong>Reputational damage and worker loyalty could also affect employers</strong></li></ul><p>New research has warned companies looking to replace human workers with AI could face major reputational consequences later on down the line if their AI strategies fall apart.</p><p>Companies that are forced to re-hire human workers after laying them off could ultimately find themselves having to pay higher salaries, offer better contractual terms and include stronger employment protections, the report from JobAssist warns.</p><p>But even after offering workers better terms overall, they could have already made reputational and loyalty damages – nearly one-third (29%) of workers say they'd return to their old job, but would immediately start looking for another.</p><h2 id="replacing-humans-with-ai-could-have-long-lasting-consequences">Replacing humans with AI could have long-lasting consequences</h2><p>While a relatively low but not insignificant number of workers would likely continue to look for other opportunities, the number of workers looking for a higher salary (71%) is much higher. Nearly two in three (64%) also said they'd negotiate more aggressively in general when returning to an old job.</p><p>However, workers are still very clearly worried about AI's impacts on their roles, because half (54%) would now prioritize job security over titles and perks, implying they're concerned about experiencing yet another redundancy. Nearly half (44%) also said they'd trust their employer less after AI replaced human workers.</p><p>All in all, the study is proof that employer reputation could suffer beyond the original workforce reduction, with laid off workers and re-hires less likely to trust the company. </p><p>And even after all the costs associated with laying off workers in the first place, and investing so heavily in artificial intelligence, JobAssist warns that workers could end up demanding more money, benefits and security than they had done before.</p><p>"For businesses evaluating the financial benefits of replacing employees with AI, the survey highlights a risk that could emerge if an automation decision has to be reversed," the report summarized.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Why a former OpenAI alumnus is building Mac data centres for AI agents ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Mount Thor secures $45 million across seed and Series A financing rounds</strong></li><li><strong>The startup deployed its first data center within six months</strong></li><li><strong>AI agents increasingly require access to familiar desktop computing environments</strong></li></ul><p>Mount Thor, a startup that runs racks of Apple computers for artificial intelligence work, has secured $45 million from investors across two funding rounds.</p><p>The company <a href="https://blog.mountthor.com/a" target="_blank" rel="nofollow">says</a> AI agents increasingly need the software environments people use, particularly when their tasks depend on applications built for macOS.</p><p>Its service makes those environments available remotely, sparing customers the expense and operational work involved in maintaining extensive Mac fleets.</p><h2 id="why-ai-agents-need-mac-infrastructure">Why AI agents need Mac infrastructure</h2><p>AI agents perform tasks through computer interfaces, sometimes operating independently for extended periods or handling requests as they arise.</p><p>Their assignments can involve navigating applications, managing files, testing software, and completing operations that require a functioning desktop environment.</p><p>When tasks depend on Apple applications or macOS-specific features, conventional <a href="https://www.techradar.com/news/best-cloud-hosting-providers">cloud servers</a> may not provide the environment developers require.</p><p>Mount Thor intends to supply remotely accessible Mac systems for research teams and businesses developing, testing, and operating agents.</p><p>Research laboratories could use its infrastructure to create varied computer environments for evaluating models that learn to control software through visual interfaces.</p><p>Companies developing agent platforms could provide separate macOS workspaces, including systems that remain active for weeks while agents perform assigned tasks.</p><p>This could allow agents to work across users' Apple devices and services through one account, subject to the necessary access permissions.</p><p>Software developers could also use the machines to build applications, check compatibility, and simulate programs across Apple's product ecosystem.</p><p>“Making Apple’s operating system and silicon available at datacenter-scale is a massive untapped opportunity in AI infrastructure,” said Peter Voell, Mount Thor CEO.</p><p>Voell argues that Apple's long investment in accessibility and human-computer interaction makes macOS particularly suitable for agents requiring human supervision.</p><p>He also considers newer Apple processors increasingly capable of handling production inference while offering improved energy efficiency and lower operating costs.</p><h2 id="funding-will-expand-capacity-and-support-inference-research">Funding will expand capacity and support inference research</h2><p>CRV and Conviction jointly led Mount Thor's Series A financing, while Chad Byers and Offline Ventures co-led its earlier seed round. CSC Leasing, Enea Capital, Long Journey Ventures, and Upper90 also invested, alongside individual backers connected with several technology companies.</p><p>Mount Thor additionally secured financing that does not involve issuing equity, providing further support for purchasing the hardware needed for expansion.</p><p>The company says it deployed its first <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> in under six months and plans to increase capacity as customer demand grows.</p><p>It intends to spend the new capital on additional computing infrastructure and recruitment, particularly for research involving open-source AI models.</p><p>Mount Thor plans to establish a laboratory for testing open-source models adapted to Apple processors, potentially reducing costs for certain workloads.</p><p>The founding team previously worked on computing infrastructure at Cerebras, CoreWeave, Fluidstack, Nebius, OpenAI, and xAI.</p><p>Brittany Walker, a general partner at CRV, said demand is increasing for computing systems capable of supporting autonomous agents.</p><p>“Mount Thor sits at the intersection of two important shifts: the rapidly expanding capabilities of autonomous agents,” Walker said.</p><p>She claims the company wants to make Apple's platforms easier to deploy at scale as businesses adopt agent-based systems.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/making-apples-operating-system-and-silicon-available-at-datacenter-scale-is-a-massive-untapped-opportunity-in-ai-infrastructure-why-a-former-openai-alumnus-is-building-mac-data-centres-for-ai-agents</link>
                                                                            <description>
                            <![CDATA[ Mount Thor raises $45 million to expand remotely accessible Apple computing infrastructure for AI agents, research laboratories, and software developers. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">nuYmyRSYNuoGoH79cdNnQ7</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/Rw3hh4WnYgWwq3Vwt4jScC-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 01:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/Rw3hh4WnYgWwq3Vwt4jScC-1920-80.jpg">
                                                            <media:credit><![CDATA[Apple]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Apple Mac Studio]]></media:description>                                                            <media:text><![CDATA[Apple Mac Studio]]></media:text>
                                <media:title type="plain"><![CDATA[Apple Mac Studio]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/Rw3hh4WnYgWwq3Vwt4jScC-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Mount Thor secures $45 million across seed and Series A financing rounds</strong></li><li><strong>The startup deployed its first data center within six months</strong></li><li><strong>AI agents increasingly require access to familiar desktop computing environments</strong></li></ul><p>Mount Thor, a startup that runs racks of Apple computers for artificial intelligence work, has secured $45 million from investors across two funding rounds.</p><p>The company <a href="https://blog.mountthor.com/a" target="_blank" rel="nofollow">says</a> AI agents increasingly need the software environments people use, particularly when their tasks depend on applications built for macOS.</p><p>Its service makes those environments available remotely, sparing customers the expense and operational work involved in maintaining extensive Mac fleets.</p><h2 id="why-ai-agents-need-mac-infrastructure">Why AI agents need Mac infrastructure</h2><p>AI agents perform tasks through computer interfaces, sometimes operating independently for extended periods or handling requests as they arise.</p><p>Their assignments can involve navigating applications, managing files, testing software, and completing operations that require a functioning desktop environment.</p><p>When tasks depend on Apple applications or macOS-specific features, conventional <a href="https://www.techradar.com/news/best-cloud-hosting-providers">cloud servers</a> may not provide the environment developers require.</p><p>Mount Thor intends to supply remotely accessible Mac systems for research teams and businesses developing, testing, and operating agents.</p><p>Research laboratories could use its infrastructure to create varied computer environments for evaluating models that learn to control software through visual interfaces.</p><p>Companies developing agent platforms could provide separate macOS workspaces, including systems that remain active for weeks while agents perform assigned tasks.</p><p>This could allow agents to work across users' Apple devices and services through one account, subject to the necessary access permissions.</p><p>Software developers could also use the machines to build applications, check compatibility, and simulate programs across Apple's product ecosystem.</p><p>“Making Apple’s operating system and silicon available at datacenter-scale is a massive untapped opportunity in AI infrastructure,” said Peter Voell, Mount Thor CEO.</p><p>Voell argues that Apple's long investment in accessibility and human-computer interaction makes macOS particularly suitable for agents requiring human supervision.</p><p>He also considers newer Apple processors increasingly capable of handling production inference while offering improved energy efficiency and lower operating costs.</p><h2 id="funding-will-expand-capacity-and-support-inference-research">Funding will expand capacity and support inference research</h2><p>CRV and Conviction jointly led Mount Thor's Series A financing, while Chad Byers and Offline Ventures co-led its earlier seed round. CSC Leasing, Enea Capital, Long Journey Ventures, and Upper90 also invested, alongside individual backers connected with several technology companies.</p><p>Mount Thor additionally secured financing that does not involve issuing equity, providing further support for purchasing the hardware needed for expansion.</p><p>The company says it deployed its first <a href="https://www.techradar.com/pro/best-data-center-proxies">data center</a> in under six months and plans to increase capacity as customer demand grows.</p><p>It intends to spend the new capital on additional computing infrastructure and recruitment, particularly for research involving open-source AI models.</p><p>Mount Thor plans to establish a laboratory for testing open-source models adapted to Apple processors, potentially reducing costs for certain workloads.</p><p>The founding team previously worked on computing infrastructure at Cerebras, CoreWeave, Fluidstack, Nebius, OpenAI, and xAI.</p><p>Brittany Walker, a general partner at CRV, said demand is increasing for computing systems capable of supporting autonomous agents.</p><p>“Mount Thor sits at the intersection of two important shifts: the rapidly expanding capabilities of autonomous agents,” Walker said.</p><p>She claims the company wants to make Apple's platforms easier to deploy at scale as businesses adopt agent-based systems.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Shipping giant wants to convert a fishing trawler into a 200-ft unmanned Arctic Russian submarine hunter for the British Royal Navy ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Kongsberg wants to convert commercial fishing trawlers into autonomous submarine hunters</strong></li><li><strong>The proposed Arctic Hunter vessels could operate independently for 60 days</strong></li><li><strong>Each 56-meter vessel would carry submarine detection sonar supplied by Thales UK</strong></li></ul><p>Norwegian maritime engineering company Kongsberg Maritime wants to convert commercial fishing trawlers into autonomous vessels capable of hunting Russian submarines for Britain's Royal Navy.</p><p>Its Arctic Hunter concept uses an existing 56-meter fishing vessel design, approximately 184 feet long, modified to conduct underwater surveillance without carrying a crew.</p><p>The proposed conversion would replace fishing equipment with autonomous navigation systems, submarine detection sonar, and secure communications equipment for extended maritime operations.</p><h2 id="british-firm-to-supply-sonar-equipment">British firm to supply sonar equipment</h2><p>Kongsberg says each vessel would carry sonar equipment supplied by Thales UK alongside protected communications technology.</p><p>The ships are expected to operate independently for approximately 60 days, allowing them to undertake extended assignments without requiring onboard personnel.</p><p>Kongsberg also intends to modify certain components, including the propulsion system, to minimize unwanted underwater noise that might interfere with acoustic monitoring equipment.</p><p>The proposed design falls under Britain's Type 92 program, which seeks relatively affordable autonomous vessels capable of assisting conventional warships with anti-submarine warfare assignments.</p><p>The concept draws inspiration from World War II, when Britain converted commercial whaling vessels into naval escorts to protect shipping against German submarines.</p><p>Kongsberg Maritime argues that commercial shipbuilding already offers suitable vessel designs for demanding North Atlantic conditions, reducing the need for completely new military construction.</p><p>The company describes its approach as a maritime engineering exercise rather than conventional weapons development.</p><h2 id="ten-vessels-planned-for-arctic-operations">Ten vessels planned for Arctic operations</h2><p>Kongsberg's proposed deployment arrangement involves 10 autonomous vessels operating from a support location west of the Bear Gap, between Norway and Svalbard.</p><p>That maritime passage connects northern waters with the Atlantic, making it an important geographical reference in discussions about Russian submarine movements.</p><p>Establishing the vessels' support facilities closer to their intended operating area would reduce the distance they must travel before beginning assignments.</p><p>But Kongsberg Maritime says it would manufacture the proposed Type 92 vessels within Britain if selected to deliver the program.</p><p>The Arctic Hunter ships would operate alongside Britain's future Type 26 frigates, which BAE Systems is constructing for several allied naval forces.</p><p>Those warships are intended for the United Kingdom, Australia, Canada, and Norway, with autonomous vessels potentially providing additional surveillance support.</p><p>The arrangement fits Britain's plans for a hybrid naval fleet combining conventional warships with remotely operated or independently navigating surface vessels.</p><p>However, Kongsberg has only outlined its proposed design, and there is no official confirmation that Britain has awarded it a construction contract.</p><p>Via <a href="https://www.defensenews.com/global/europe/2026/11/09/firm-bets-on-converted-fishing-trawlers-to-tackle-russian-submarines-in-the-arctic/" target="_blank" rel="nofollow">Defense News</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/shipping-giant-wants-to-convert-a-fishing-trawler-into-a-200-ft-unmanned-arctic-russian-submarine-hunter-for-the-british-royal-navy</link>
                                                                            <description>
                            <![CDATA[ Kongsberg Maritime proposes converting 56-meter fishing trawlers into autonomous Arctic submarine hunters, with 10 vessels potentially supporting Britain's Royal Navy. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">mQCrqG37VdKa9q3tm3UK5n</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/69oaGNAbFCupFf8bW4MxSD-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 00:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/69oaGNAbFCupFf8bW4MxSD-1920-80.jpg">
                                                            <media:credit><![CDATA[Future | Joshua Chard]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Looking out from a boat on a Nile river to the shores]]></media:description>                                                            <media:text><![CDATA[Looking out from a boat on a Nile river to the shores]]></media:text>
                                <media:title type="plain"><![CDATA[Looking out from a boat on a Nile river to the shores]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/69oaGNAbFCupFf8bW4MxSD-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Kongsberg wants to convert commercial fishing trawlers into autonomous submarine hunters</strong></li><li><strong>The proposed Arctic Hunter vessels could operate independently for 60 days</strong></li><li><strong>Each 56-meter vessel would carry submarine detection sonar supplied by Thales UK</strong></li></ul><p>Norwegian maritime engineering company Kongsberg Maritime wants to convert commercial fishing trawlers into autonomous vessels capable of hunting Russian submarines for Britain's Royal Navy.</p><p>Its Arctic Hunter concept uses an existing 56-meter fishing vessel design, approximately 184 feet long, modified to conduct underwater surveillance without carrying a crew.</p><p>The proposed conversion would replace fishing equipment with autonomous navigation systems, submarine detection sonar, and secure communications equipment for extended maritime operations.</p><h2 id="british-firm-to-supply-sonar-equipment">British firm to supply sonar equipment</h2><p>Kongsberg says each vessel would carry sonar equipment supplied by Thales UK alongside protected communications technology.</p><p>The ships are expected to operate independently for approximately 60 days, allowing them to undertake extended assignments without requiring onboard personnel.</p><p>Kongsberg also intends to modify certain components, including the propulsion system, to minimize unwanted underwater noise that might interfere with acoustic monitoring equipment.</p><p>The proposed design falls under Britain's Type 92 program, which seeks relatively affordable autonomous vessels capable of assisting conventional warships with anti-submarine warfare assignments.</p><p>The concept draws inspiration from World War II, when Britain converted commercial whaling vessels into naval escorts to protect shipping against German submarines.</p><p>Kongsberg Maritime argues that commercial shipbuilding already offers suitable vessel designs for demanding North Atlantic conditions, reducing the need for completely new military construction.</p><p>The company describes its approach as a maritime engineering exercise rather than conventional weapons development.</p><h2 id="ten-vessels-planned-for-arctic-operations">Ten vessels planned for Arctic operations</h2><p>Kongsberg's proposed deployment arrangement involves 10 autonomous vessels operating from a support location west of the Bear Gap, between Norway and Svalbard.</p><p>That maritime passage connects northern waters with the Atlantic, making it an important geographical reference in discussions about Russian submarine movements.</p><p>Establishing the vessels' support facilities closer to their intended operating area would reduce the distance they must travel before beginning assignments.</p><p>But Kongsberg Maritime says it would manufacture the proposed Type 92 vessels within Britain if selected to deliver the program.</p><p>The Arctic Hunter ships would operate alongside Britain's future Type 26 frigates, which BAE Systems is constructing for several allied naval forces.</p><p>Those warships are intended for the United Kingdom, Australia, Canada, and Norway, with autonomous vessels potentially providing additional surveillance support.</p><p>The arrangement fits Britain's plans for a hybrid naval fleet combining conventional warships with remotely operated or independently navigating surface vessels.</p><p>However, Kongsberg has only outlined its proposed design, and there is no official confirmation that Britain has awarded it a construction contract.</p><p>Via <a href="https://www.defensenews.com/global/europe/2026/11/09/firm-bets-on-converted-fishing-trawlers-to-tackle-russian-submarines-in-the-arctic/" target="_blank" rel="nofollow">Defense News</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ A major breakthrough has claimed to set a world record for encoding data onto a quantum computer — and, crucially, reading it back again ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Moth Quantum retrieves thousands of stored values from IBM quantum hardware</strong></li><li><strong>IBM’s Phoenix processor successfully retrieves information distributed across all 120 qubits</strong></li><li><strong>Quantum hardware struggles to reproduce complex images with sufficient accuracy</strong></li></ul><p>Moth Quantum says its Tessa technique has achieved a new quantum computing record by storing information on a processor and successfully retrieving it.</p><p>The method placed 1,568 numerical values across IBM’s 120-qubit Phoenix processor, then used the recovered information to recreate an image.</p><p>This addresses a major challenge in quantum computing: getting ordinary information into quantum machines and retrieving it without significant loss of accuracy.</p><h2 id="a-new-record-with-important-accuracy-limits">A new record with important accuracy limits</h2><p>Previous research of this kind recovered 128 numerical values using 12 trapped-ion qubits, making the latest result substantially larger by the study’s definition.</p><p>Tessa uses neighboring qubits to control where information is stored, allowing the processor’s existing connections to handle the operation without moving information between distant qubits.</p><p>The arrangement avoids extra qubit-moving operations and additional helper qubits, while the tested circuits reached a depth of 19 two-qubit operations.</p><p>“Every quantum computer has to get data in before it can do anything useful with it, and that step has been a bottleneck for fields like quantum machine learning,” said Daniel Bultrini, the Moth researcher who developed Tessa.</p><p>“Tessa makes the most of the chip’s capabilities, so the whole processor can be put to work at once.”</p><p>Image tests stored 784 and 441 values, requiring 728 and 428 controlled operations, respectively; each used 8,192 measurement attempts.</p><p>However, information recovery was imperfect: the largest color image achieved a correlation of 0.69 with its source, compared with 0.95–0.96 for simpler images.</p><p>The average numerical error was 0.130 on Phoenix and 0.131 on IBM’s Fez processor, falling to 0.075 and 0.078 after adjustment.</p><p>That adjustment used two statistics taken from the original image, meaning the improved reconstruction relied partly on information already known.</p><p>For a separate test combining image brightness and color, Phoenix recorded errors of 0.152 and 0.092, while Fez recorded 0.144 and 0.084.</p><p>The Phoenix brightness error was 2.6 times higher than the company’s noise model predicted, indicating that real hardware introduced additional inaccuracies.</p><h2 id="audio-and-biological-tests">Audio and biological tests</h2><p>Moth Quantum also tried audio, musical notes, DNA, and protein sequences, using different ways to represent information inside the processor.</p><p>A music test containing 347 flute notes recovered 18.7% of pitches correctly before adjustment, improving to 49.3% afterward on the Fez processor.</p><p>Biological tests required 32,768 measurement attempts, compared with 8,192 for image experiments, yet accuracy varied substantially between sequences for every measurement basis.</p><p>The p53 protein sequence achieved 71.5% correct recovery, while Sonic hedgehog reached 76.2% and mitochondrial protein MT-CO3 recorded 92.3% on IBM’s Phoenix hardware.</p><p>Results were weaker for other samples: SUR1 reached 22.6%, while pikachurin achieved 25.8%, showing that longer or more complex sequences remained difficult.</p><p>The tests also included a mitochondrial DNA sequence containing 784 building blocks and a POLG2 messenger RNA sequence with 1,582.</p><p>The study warns that one incorrect DNA or protein unit can matter biologically, so these recovery percentages do not establish dependable medical applications.</p><p>The simulations here do not include all interference between neighboring processor groups, and the hardware experiments used no error-correction techniques.</p><p>Tessa also depends on a particular pattern of connections between qubits, while measurement demands rise sharply as those connections become more numerous.</p><p>Moth Quantum did not use error mitigation, which the study says could reduce reconstruction errors significantly in future hardware tests.</p><p>“The quantum applications that will have the greatest impact, including those that work closely with AI, will depend on breakthroughs such as Tessa in order to meet the expectations that have been set,” said Harry Kumar, Moth’s Founder and CCO.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/a-major-breakthrough-has-claimed-to-set-a-world-record-for-encoding-data-onto-a-quantum-computer-and-crucially-reading-it-back-again</link>
                                                                            <description>
                            <![CDATA[ Moth Quantum’s Tessa technique sets a quantum data encoding record, although image reconstruction errors and biological testing show some limitations. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">ycYrSDbZ2pxy7ssksUKtpj</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/d4oN2QTeNf8QYJDmjZnAKE-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Mon, 16 Nov 2026 00:01:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/d4oN2QTeNf8QYJDmjZnAKE-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Quantum computing]]></media:description>                                                            <media:text><![CDATA[Quantum computing]]></media:text>
                                <media:title type="plain"><![CDATA[Quantum computing]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/d4oN2QTeNf8QYJDmjZnAKE-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Moth Quantum retrieves thousands of stored values from IBM quantum hardware</strong></li><li><strong>IBM’s Phoenix processor successfully retrieves information distributed across all 120 qubits</strong></li><li><strong>Quantum hardware struggles to reproduce complex images with sufficient accuracy</strong></li></ul><p>Moth Quantum says its Tessa technique has achieved a new quantum computing record by storing information on a processor and successfully retrieving it.</p><p>The method placed 1,568 numerical values across IBM’s 120-qubit Phoenix processor, then used the recovered information to recreate an image.</p><p>This addresses a major challenge in quantum computing: getting ordinary information into quantum machines and retrieving it without significant loss of accuracy.</p><h2 id="a-new-record-with-important-accuracy-limits">A new record with important accuracy limits</h2><p>Previous research of this kind recovered 128 numerical values using 12 trapped-ion qubits, making the latest result substantially larger by the study’s definition.</p><p>Tessa uses neighboring qubits to control where information is stored, allowing the processor’s existing connections to handle the operation without moving information between distant qubits.</p><p>The arrangement avoids extra qubit-moving operations and additional helper qubits, while the tested circuits reached a depth of 19 two-qubit operations.</p><p>“Every quantum computer has to get data in before it can do anything useful with it, and that step has been a bottleneck for fields like quantum machine learning,” said Daniel Bultrini, the Moth researcher who developed Tessa.</p><p>“Tessa makes the most of the chip’s capabilities, so the whole processor can be put to work at once.”</p><p>Image tests stored 784 and 441 values, requiring 728 and 428 controlled operations, respectively; each used 8,192 measurement attempts.</p><p>However, information recovery was imperfect: the largest color image achieved a correlation of 0.69 with its source, compared with 0.95–0.96 for simpler images.</p><p>The average numerical error was 0.130 on Phoenix and 0.131 on IBM’s Fez processor, falling to 0.075 and 0.078 after adjustment.</p><p>That adjustment used two statistics taken from the original image, meaning the improved reconstruction relied partly on information already known.</p><p>For a separate test combining image brightness and color, Phoenix recorded errors of 0.152 and 0.092, while Fez recorded 0.144 and 0.084.</p><p>The Phoenix brightness error was 2.6 times higher than the company’s noise model predicted, indicating that real hardware introduced additional inaccuracies.</p><h2 id="audio-and-biological-tests">Audio and biological tests</h2><p>Moth Quantum also tried audio, musical notes, DNA, and protein sequences, using different ways to represent information inside the processor.</p><p>A music test containing 347 flute notes recovered 18.7% of pitches correctly before adjustment, improving to 49.3% afterward on the Fez processor.</p><p>Biological tests required 32,768 measurement attempts, compared with 8,192 for image experiments, yet accuracy varied substantially between sequences for every measurement basis.</p><p>The p53 protein sequence achieved 71.5% correct recovery, while Sonic hedgehog reached 76.2% and mitochondrial protein MT-CO3 recorded 92.3% on IBM’s Phoenix hardware.</p><p>Results were weaker for other samples: SUR1 reached 22.6%, while pikachurin achieved 25.8%, showing that longer or more complex sequences remained difficult.</p><p>The tests also included a mitochondrial DNA sequence containing 784 building blocks and a POLG2 messenger RNA sequence with 1,582.</p><p>The study warns that one incorrect DNA or protein unit can matter biologically, so these recovery percentages do not establish dependable medical applications.</p><p>The simulations here do not include all interference between neighboring processor groups, and the hardware experiments used no error-correction techniques.</p><p>Tessa also depends on a particular pattern of connections between qubits, while measurement demands rise sharply as those connections become more numerous.</p><p>Moth Quantum did not use error mitigation, which the study says could reduce reconstruction errors significantly in future hardware tests.</p><p>“The quantum applications that will have the greatest impact, including those that work closely with AI, will depend on breakthroughs such as Tessa in order to meet the expectations that have been set,” said Harry Kumar, Moth’s Founder and CCO.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ KTC heads to CES 2027 with a 31.5-inch 6K Mini-LED display engineered to break records ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>KTC is preparing a 31.5-inch 6K Mini-LED monitor for CES 2027</strong></li><li><strong>The KTC M32P2 6K monitor delivers 6,016 x 3,384 resolution at 60Hz</strong></li><li><strong>Its Mini-LED backlight features 2,304 local dimming zones</strong></li></ul><p>KTC is preparing to challenge the expensive end of the high-resolution monitor market with a new display scheduled for CES 2027.</p><p>The upcoming M32P2 combines a 31.5-inch IPS panel with 6K resolution and Mini-LED backlighting, giving professionals another option.</p><p>Although pricing remains undisclosed, KTC's history of selling advanced displays at relatively accessible prices makes this announcement particularly interesting.</p><h2 id="ktc-brings-6k-resolution-and-2-304-dimming-zones-to-ces">KTC brings 6K resolution and 2,304 dimming zones to CES</h2><p>The KTC M32P2 6K <a href="https://www.techradar.com/news/computing-components/peripherals/best-monitor-9-reviewed-and-rated-1058662">monitor</a> sports a 6K panel delivering 6,016 x 3,384 pixels at 60Hz, offering substantially more workspace than conventional 4K displays.</p><p>The Mini-LED backlight also incorporates 2,304 individually controlled dimming zones, allowing different screen areas to receive different levels of illumination.</p><p>KTC specifies peak brightness of 1,000 nits when displaying HDR content, while standard viewing conditions get a maximum rating of 400 nits.</p><p>The panel has a typical contrast ratio of 1,000:1 and supports approximately 1.07 billion colors through its 8-bit plus FRC implementation.</p><p>Its advertised color reproduction figures include 98% Adobe RGB coverage, 99% DCI-P3 coverage, and complete 100% sRGB coverage.</p><p>These specifications could make the display relevant to professional image editing and content production, although independent testing will determine actual color accuracy.</p><p>The company has also included connectivity options covering two HDMI 2.1 inputs and one DisplayPort 1.4 connection, along with a USB-C port supporting the full display resolution while delivering up to 90W of power to connected devices.</p><p>Additional connections include three USB 3.0 ports, divided between one upstream and two downstream connections, alongside a headphone output.</p><p>The adjustable stand provides between -5 and 20 degrees of tilt, alongside 110mm of height adjustment for different working arrangements.</p><p>Buyers can also use its 100 x 100mm VESA mounting points when installing the display on compatible arms or wall mounts.</p><p>KTC acknowledged that the high-resolution category remains relatively limited, describing its upcoming exhibition units as unfinished products.</p><p>“We don’t have many 5K or 6K monitors available at the moment. We have these two units at the show, but they are only samples and not launched yet,” a marketing specialist from KTC told TechRadar Pro.</p><h2 id="ktc-39-s-pricing-history-could-make-its-6k-monitor-interesting">KTC's pricing history could make its 6K monitor interesting</h2><p>The biggest unanswered question surrounding the M32P2 concerns its eventual retail price, particularly considering KTC's established reputation for affordable display hardware.</p><p>Its H27P3 27-inch <a href="https://www.techradar.com/best/best-5k-and-8k-monitors">5K monitor</a> has previously sold for approximately $530, considerably less than several competing professional displays.</p><p>The company has also attracted attention with affordable Mini-LED <a href="https://www.techradar.com/news/best-gaming-monitor">gaming monitors</a>, including its M27T6S, which has appeared at around $300.</p><p>Those products provide some indication of KTC's pricing approach when introducing display technologies normally associated with more expensive products.</p><p>However, neither example establishes what the manufacturer will charge for its upcoming 31.5-inch model. </p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/exclusive-ktc-heads-to-ces-2027-with-a-31-5-inch-6k-mini-led-display-engineered-to-break-records-as-the-cheapest-ultra-high-resolution-monitor-ever-sold</link>
                                                                            <description>
                            <![CDATA[ KTC will showcase its 31.5-inch 6K Mini-LED monitor at CES 2027, featuring 2,304 dimming zones and professional color reproduction. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">AtpKvCXeUDrhox82j5nPn</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/JN3uP2FdbfCn926VD4EBmQ-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Sun, 15 Nov 2026 20:35:00 +0000</pubDate>                                                                                                                                <updated>Sun, 15 Nov 2026 22:13:07 +0000</updated>
                                                                                                                                            <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/JN3uP2FdbfCn926VD4EBmQ-1920-80.png">
                                                            <media:credit><![CDATA[KTC]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[KTC M32P2 6K Monitor set to arrive at CES 2027 in Las Vegas]]></media:description>                                                            <media:text><![CDATA[KTC M32P2 6K Monitor set to arrive at CES 2027 in Las Vegas]]></media:text>
                                <media:title type="plain"><![CDATA[KTC M32P2 6K Monitor set to arrive at CES 2027 in Las Vegas]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/JN3uP2FdbfCn926VD4EBmQ-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>KTC is preparing a 31.5-inch 6K Mini-LED monitor for CES 2027</strong></li><li><strong>The KTC M32P2 6K monitor delivers 6,016 x 3,384 resolution at 60Hz</strong></li><li><strong>Its Mini-LED backlight features 2,304 local dimming zones</strong></li></ul><p>KTC is preparing to challenge the expensive end of the high-resolution monitor market with a new display scheduled for CES 2027.</p><p>The upcoming M32P2 combines a 31.5-inch IPS panel with 6K resolution and Mini-LED backlighting, giving professionals another option.</p><p>Although pricing remains undisclosed, KTC's history of selling advanced displays at relatively accessible prices makes this announcement particularly interesting.</p><h2 id="ktc-brings-6k-resolution-and-2-304-dimming-zones-to-ces">KTC brings 6K resolution and 2,304 dimming zones to CES</h2><p>The KTC M32P2 6K <a href="https://www.techradar.com/news/computing-components/peripherals/best-monitor-9-reviewed-and-rated-1058662">monitor</a> sports a 6K panel delivering 6,016 x 3,384 pixels at 60Hz, offering substantially more workspace than conventional 4K displays.</p><p>The Mini-LED backlight also incorporates 2,304 individually controlled dimming zones, allowing different screen areas to receive different levels of illumination.</p><p>KTC specifies peak brightness of 1,000 nits when displaying HDR content, while standard viewing conditions get a maximum rating of 400 nits.</p><p>The panel has a typical contrast ratio of 1,000:1 and supports approximately 1.07 billion colors through its 8-bit plus FRC implementation.</p><p>Its advertised color reproduction figures include 98% Adobe RGB coverage, 99% DCI-P3 coverage, and complete 100% sRGB coverage.</p><p>These specifications could make the display relevant to professional image editing and content production, although independent testing will determine actual color accuracy.</p><p>The company has also included connectivity options covering two HDMI 2.1 inputs and one DisplayPort 1.4 connection, along with a USB-C port supporting the full display resolution while delivering up to 90W of power to connected devices.</p><p>Additional connections include three USB 3.0 ports, divided between one upstream and two downstream connections, alongside a headphone output.</p><p>The adjustable stand provides between -5 and 20 degrees of tilt, alongside 110mm of height adjustment for different working arrangements.</p><p>Buyers can also use its 100 x 100mm VESA mounting points when installing the display on compatible arms or wall mounts.</p><p>KTC acknowledged that the high-resolution category remains relatively limited, describing its upcoming exhibition units as unfinished products.</p><p>“We don’t have many 5K or 6K monitors available at the moment. We have these two units at the show, but they are only samples and not launched yet,” a marketing specialist from KTC told TechRadar Pro.</p><h2 id="ktc-39-s-pricing-history-could-make-its-6k-monitor-interesting">KTC's pricing history could make its 6K monitor interesting</h2><p>The biggest unanswered question surrounding the M32P2 concerns its eventual retail price, particularly considering KTC's established reputation for affordable display hardware.</p><p>Its H27P3 27-inch <a href="https://www.techradar.com/best/best-5k-and-8k-monitors">5K monitor</a> has previously sold for approximately $530, considerably less than several competing professional displays.</p><p>The company has also attracted attention with affordable Mini-LED <a href="https://www.techradar.com/news/best-gaming-monitor">gaming monitors</a>, including its M27T6S, which has appeared at around $300.</p><p>Those products provide some indication of KTC's pricing approach when introducing display technologies normally associated with more expensive products.</p><p>However, neither example establishes what the manufacturer will charge for its upcoming 31.5-inch model. </p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Your next desktop PC for the price of a Ferrari? HP ZGX Fury packs in 748GB RAM and Nvidia GB300 — but prices start at $132,000 ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>HP's latest workstation packs 748GB of memory and Nvidia's GB300 processor</strong></li><li><strong>The ZGX Fury starts at an eye-watering $132,000</strong></li><li><strong>LPDDR5X and HBM3e memory combine to provide 748GB total capacity</strong></li></ul><p>HP has introduced an expensive workstation that brings computing hardware previously associated with large-scale AI infrastructure into an office-sized machine.</p><p>The ZGX Fury G1n combines Nvidia's GB300 Grace Blackwell Ultra processor with an enormous memory allocation designed for demanding artificial intelligence workloads.</p><p>Its starting price of $132,000 makes it considerably more expensive than conventional professional computers, even before customers consider optional upgrades</p><h2 id="hp-crams-748gb-memory-into-its-expensive-workstation">HP crams 748GB memory into its expensive workstation</h2><p>The <a href="http://techradar.com/news/best-mobile-workstations">mobile workstation</a> combines two different memory technologies, LPDDR5X and HBM3e, with 496GB allocated to system memory and another 252GB dedicated to graphics processing.</p><p>The LPDDR5X allocation runs at 6,400 MT/s and connects to a 72-core Arm Neoverse V2 Grace CPU, supporting demanding processing workloads.</p><p>Meanwhile, the HBM3e memory is integrated into Nvidia's Blackwell Ultra GPU, providing dedicated high-bandwidth capacity for graphics and AI operations.</p><p>Customers opting for additional graphics hardware can increase the available video memory to 276GB through selected configurations.</p><p>Together, these components provide the advertised 748GB capacity, although customers purchasing additional graphics hardware can increase available video memory to 276GB.</p><p>HP also offers configurations featuring discrete Nvidia RTX PRO <a href="https://www.techradar.com/best/best-workstations">workstation</a> graphics cards, including the RTX PRO 2000 and RTX PRO 4000.</p><p>The memory capacity serves a particular purpose beyond running conventional office applications or demanding professional software.</p><p>Developers and researchers working with <a href="https://www.techradar.com/computing/artificial-intelligence/best-llms">large language models</a> can use the machine to process substantial workloads without depending entirely on remote computing infrastructure.</p><p>HP says the available memory can accommodate local adjustment of language models containing up to 100 billion parameters.</p><p>The system can also perform inference using models reaching one trillion parameters through FP4 quantization, although actual performance depends on workload requirements.</p><h2 id="liquid-cooling-and-networking-accompany-the-132-000-price">Liquid cooling and networking accompany the $132,000 price</h2><p>Running powerful computing components inside a relatively compact enclosure introduces thermal management requirements that ordinary desktop cooling systems cannot necessarily accommodate.</p><p>HP therefore equips the workstation with liquid cooling alongside a 1,600W Titanium-certified power supply.</p><p>The device ships with Ubuntu 24.04 LTS, making Linux its supplied operating system rather than Microsoft's Windows environment.</p><p>This configuration caters to developers and technical professionals who need software environments suitable for demanding computing and AI development.</p><p>Networking is another important component, with an Nvidia ConnectX-8 SuperNIC providing two QSFP112 connections rated at 400Gbps.</p><p>These connections allow multiple machines to communicate through high-speed networking when organizations need additional computing resources.</p><p>The enclosure also supports installation inside a 5U equipment rack, giving organizations another deployment option beyond placing it beneath an office desk.</p><p>This flexibility allows organizations to deploy the workstation individually or install multiple units inside equipment racks for larger computing operations.</p><p>Getting all these specifications will cost $132,000 for the standard configuration, enough to buy a brand-new 2026 Cadillac Escalade IQL, although not quite a Ferrari.</p><p>Customers opting for the RTX PRO 2000 configuration will pay $133,000, while the more powerful RTX PRO 4000 option raises the price to $136,000.</p><p>At these prices, the ZGX Fury is unlikely to attract ordinary desktop buyers looking for additional memory or improved everyday computing performance.</p><p>Instead, its potential customers are organizations that require substantial local computing resources and can justify spending considerably more than conventional workstation prices.</p><p>Via <a href="https://hothardware.com/news/hp-zgx-fury-packs-748gb-ram-and-nvidia-gb300-starting-at-132000" target="_blank" rel="nofollow">Hothardware</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/your-next-desktop-pc-for-the-price-of-a-ferrari-hp-zgx-fury-packs-in-748gb-ram-and-nvidia-gb300-but-starts-at-usd132-000</link>
                                                                            <description>
                            <![CDATA[ HP's $132,000 ZGX Fury combines Nvidia's GB300 processor, 748GB memory, liquid cooling, and high-speed networking for demanding artificial intelligence workloads. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">LYhcbhfetfBC2cF3E63KwA</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/PZgCJ7B7RendGrosbZtv5e-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sun, 15 Nov 2026 19:15:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/PZgCJ7B7RendGrosbZtv5e-1920-80.jpg">
                                                            <media:credit><![CDATA[Videocardz]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[HP ZGX Fury desktop workstation]]></media:description>                                                            <media:text><![CDATA[HP ZGX Fury desktop workstation]]></media:text>
                                <media:title type="plain"><![CDATA[HP ZGX Fury desktop workstation]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/PZgCJ7B7RendGrosbZtv5e-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>HP's latest workstation packs 748GB of memory and Nvidia's GB300 processor</strong></li><li><strong>The ZGX Fury starts at an eye-watering $132,000</strong></li><li><strong>LPDDR5X and HBM3e memory combine to provide 748GB total capacity</strong></li></ul><p>HP has introduced an expensive workstation that brings computing hardware previously associated with large-scale AI infrastructure into an office-sized machine.</p><p>The ZGX Fury G1n combines Nvidia's GB300 Grace Blackwell Ultra processor with an enormous memory allocation designed for demanding artificial intelligence workloads.</p><p>Its starting price of $132,000 makes it considerably more expensive than conventional professional computers, even before customers consider optional upgrades</p><h2 id="hp-crams-748gb-memory-into-its-expensive-workstation">HP crams 748GB memory into its expensive workstation</h2><p>The <a href="http://techradar.com/news/best-mobile-workstations">mobile workstation</a> combines two different memory technologies, LPDDR5X and HBM3e, with 496GB allocated to system memory and another 252GB dedicated to graphics processing.</p><p>The LPDDR5X allocation runs at 6,400 MT/s and connects to a 72-core Arm Neoverse V2 Grace CPU, supporting demanding processing workloads.</p><p>Meanwhile, the HBM3e memory is integrated into Nvidia's Blackwell Ultra GPU, providing dedicated high-bandwidth capacity for graphics and AI operations.</p><p>Customers opting for additional graphics hardware can increase the available video memory to 276GB through selected configurations.</p><p>Together, these components provide the advertised 748GB capacity, although customers purchasing additional graphics hardware can increase available video memory to 276GB.</p><p>HP also offers configurations featuring discrete Nvidia RTX PRO <a href="https://www.techradar.com/best/best-workstations">workstation</a> graphics cards, including the RTX PRO 2000 and RTX PRO 4000.</p><p>The memory capacity serves a particular purpose beyond running conventional office applications or demanding professional software.</p><p>Developers and researchers working with <a href="https://www.techradar.com/computing/artificial-intelligence/best-llms">large language models</a> can use the machine to process substantial workloads without depending entirely on remote computing infrastructure.</p><p>HP says the available memory can accommodate local adjustment of language models containing up to 100 billion parameters.</p><p>The system can also perform inference using models reaching one trillion parameters through FP4 quantization, although actual performance depends on workload requirements.</p><h2 id="liquid-cooling-and-networking-accompany-the-132-000-price">Liquid cooling and networking accompany the $132,000 price</h2><p>Running powerful computing components inside a relatively compact enclosure introduces thermal management requirements that ordinary desktop cooling systems cannot necessarily accommodate.</p><p>HP therefore equips the workstation with liquid cooling alongside a 1,600W Titanium-certified power supply.</p><p>The device ships with Ubuntu 24.04 LTS, making Linux its supplied operating system rather than Microsoft's Windows environment.</p><p>This configuration caters to developers and technical professionals who need software environments suitable for demanding computing and AI development.</p><p>Networking is another important component, with an Nvidia ConnectX-8 SuperNIC providing two QSFP112 connections rated at 400Gbps.</p><p>These connections allow multiple machines to communicate through high-speed networking when organizations need additional computing resources.</p><p>The enclosure also supports installation inside a 5U equipment rack, giving organizations another deployment option beyond placing it beneath an office desk.</p><p>This flexibility allows organizations to deploy the workstation individually or install multiple units inside equipment racks for larger computing operations.</p><p>Getting all these specifications will cost $132,000 for the standard configuration, enough to buy a brand-new 2026 Cadillac Escalade IQL, although not quite a Ferrari.</p><p>Customers opting for the RTX PRO 2000 configuration will pay $133,000, while the more powerful RTX PRO 4000 option raises the price to $136,000.</p><p>At these prices, the ZGX Fury is unlikely to attract ordinary desktop buyers looking for additional memory or improved everyday computing performance.</p><p>Instead, its potential customers are organizations that require substantial local computing resources and can justify spending considerably more than conventional workstation prices.</p><p>Via <a href="https://hothardware.com/news/hp-zgx-fury-packs-748gb-ram-and-nvidia-gb300-starting-at-132000" target="_blank" rel="nofollow">Hothardware</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ 'It would make me think twice about applying': Workers are pushing back at companies who insist on AI interviews ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Half of Brits say they've been interviewed by an AI avatar, not a human</strong></li><li><strong>11% of UK workers say they'd probably or definitely ditch the job application</strong></li><li><strong>AI can help admin, but human connections are vital in the recruitment process</strong></li></ul><p>Seven in 10 UK workers say they'd feel uncomfortable if their interview was conducted by an AI assisted, new Careerminds research has revealed, while not even a third say they'd continue without hesitation.</p><p>But the reality is a stark contrast, because as many as half (47%) of UK job seekers have already been interviewed by AI, per a separate Greenhouse study.</p><p>But despite hesitation, where workers sit can vary quite widely. Around one-third (33%) would continue but feel uncomfortable and a quarter (25%) say it would depend on the role or salary – only 11% say they would withdraw entirely.</p><h2 id="most-workers-feel-uncomfortable-about-ai-interviews-but-only-one-in-ten-would-withdraw">Most workers feel uncomfortable about AI interviews, but only one in ten would withdraw</h2><p>Of those 11%, only 5.5% would definitely withdraw, with the other 5.7% probably considering withdrawing. Still, the figures confirm that recruiters could be losing access to vital human talent if they use AI avatars to interview candidates.</p><p>The Greenhouse study put the figure much higher, at 30% of UK candidates abandoning the hiring process because of an AI interview, stressing that four in five (82%) weren't even told beforehand that they'd be answering questions with an AI avatar.</p><p>Unsurprisingly, older generations would be more likely to withdraw, Careerminds found, highlighting the challenges that older workers could face when looking for a job change.</p><p>Resident careers expert Amanda Augustine stressed that interviews serve as a crucial first point of contact for many workers, and they can influence how that candidate goes on to perceive the employer.</p><p>One worker cited in the study described his AI interview as "surreal, strange and impersonal." He noted that the whole experience "felt like a one-way process."</p><p>Augustine said the best fit for AI is actually in helping recruiters to process the administrative side of the job, such as trawling through applications and refining the pool. "Let the technology handle the administrative tasks so your team has more time for the human conversations that help both sides decide whether the role is the right fit," she added.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/it-would-make-me-think-twice-about-applying-workers-are-pushing-back-at-companies-who-insist-on-ai-interviews</link>
                                                                            <description>
                            <![CDATA[ Half of Brits have had a job interview with an AI avatar – many feel uncomfortable, some say they'd even scrap the job application altogether. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">pSgunBLJF6hRyL3BA54bd</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/MtYFJW8w34KGWQVpbFEvoX-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sun, 15 Nov 2026 12:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/MtYFJW8w34KGWQVpbFEvoX-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock/Stokkete]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Man and robot waiting for an interview]]></media:description>                                                            <media:text><![CDATA[Man and robot waiting for an interview]]></media:text>
                                <media:title type="plain"><![CDATA[Man and robot waiting for an interview]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/MtYFJW8w34KGWQVpbFEvoX-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Half of Brits say they've been interviewed by an AI avatar, not a human</strong></li><li><strong>11% of UK workers say they'd probably or definitely ditch the job application</strong></li><li><strong>AI can help admin, but human connections are vital in the recruitment process</strong></li></ul><p>Seven in 10 UK workers say they'd feel uncomfortable if their interview was conducted by an AI assisted, new Careerminds research has revealed, while not even a third say they'd continue without hesitation.</p><p>But the reality is a stark contrast, because as many as half (47%) of UK job seekers have already been interviewed by AI, per a separate Greenhouse study.</p><p>But despite hesitation, where workers sit can vary quite widely. Around one-third (33%) would continue but feel uncomfortable and a quarter (25%) say it would depend on the role or salary – only 11% say they would withdraw entirely.</p><h2 id="most-workers-feel-uncomfortable-about-ai-interviews-but-only-one-in-ten-would-withdraw">Most workers feel uncomfortable about AI interviews, but only one in ten would withdraw</h2><p>Of those 11%, only 5.5% would definitely withdraw, with the other 5.7% probably considering withdrawing. Still, the figures confirm that recruiters could be losing access to vital human talent if they use AI avatars to interview candidates.</p><p>The Greenhouse study put the figure much higher, at 30% of UK candidates abandoning the hiring process because of an AI interview, stressing that four in five (82%) weren't even told beforehand that they'd be answering questions with an AI avatar.</p><p>Unsurprisingly, older generations would be more likely to withdraw, Careerminds found, highlighting the challenges that older workers could face when looking for a job change.</p><p>Resident careers expert Amanda Augustine stressed that interviews serve as a crucial first point of contact for many workers, and they can influence how that candidate goes on to perceive the employer.</p><p>One worker cited in the study described his AI interview as "surreal, strange and impersonal." He noted that the whole experience "felt like a one-way process."</p><p>Augustine said the best fit for AI is actually in helping recruiters to process the administrative side of the job, such as trawling through applications and refining the pool. "Let the technology handle the administrative tasks so your team has more time for the human conversations that help both sides decide whether the role is the right fit," she added.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Today I learned why wind farms paint one turbine blade black ]]></title>
                                                                                                <dc:content><![CDATA[ <p>You'd think that painting a single turbine blade black is a pretty inexpensive fix for accidental bird deaths. It sounds almost too good to be true, which perhaps it is. </p><p>At the Smøla wind farm in Norway, every wind turbine had one of its blade painted black for an extended science experiment back in August 2013. Over the next three years, the number of bird carcasses found fell by 70%. </p><p>However, similar experiments in the Netherlands did not yield positive results, because apparently this fix depends heavily on the local bird species and population size. </p><h2 id="too-many-dead-birds">Too many dead birds</h2><p>There are a total of 68 turbines at the Smøla wind power plant, an archipelago off the coast of Central Norway. Researchers from the Norwegian Institute for Nature Research compared four regular turbines with four neighboring ones that had one blade painted black for this experiment.</p><p>They regularly counted the number of bird carcasses within 100 meters of the wind farm. This continued over a period of more than ten years, combining 7.5 years of baseline data from before painting any turbine blades to 3.5 years of data from after. In 2020, they published their findings in a paper called Ecology and Evolution.</p><p>Turns out that this small change in visual cues reduced the number of bird deaths by over 70%, proving especially effective on the somewhat rare white-tailed sea eagle population of the region. While the researchers noted that this anecdotal data may not hold up to different bird populations, this was still independent real-world confirmation of a separate laboratory experiment that found similar results in 2003.</p><h2 id="how-a-black-blade-is-supposed-to-help">How a black blade is supposed to help</h2><p>What difference does one black blade make? According to researchers, it has to do with how birds process visual cues. </p><p>Animation designers call it motion smearing, where images that move too fast for the human eye to follow are inferred in the brain as one continuous motion. When turbine blades are in motion, birds find it difficult to notice the threat because the unicolor rotating blades look like one giant circle. </p><p>When you paint one of these blades black, it breaks up the pattern so the moving blades are more easily distinguishable to the eye. But as <a href="https://www.audubon.org/magazine/can-painting-wind-turbine-blades-black-really-save-birds" target="_blank" rel="nofollow">Audubon reported</a>,  these are preliminary findings that won't translate to all bird species equally — especially ones that don't have the visual acuity to notice the difference. There's also a significant amount of effort involved, since painting turbines in operation requires both lifts and specialist crews who can navigate the height. </p><h2 id="when-the-trick-stops-working">When the trick stops working</h2><p>Most of the land experiments so far have stuck with black, but at sea, a new trial is experimenting with a different color.</p><p>Dutch consortium Ecowende is fitting<a href="https://ocean-energyresources.com/2026/05/10/red-blades-real-world-test-for-nature-inclusive-offshore-wind/" target="_blank" rel="nofollow"> seven Vestas V236-15.0 MW turbines</a> with one red 115.5-meter blade each. Red was chosen after weighing options against criteria such as durability and heat management, apart from the obvious benefit of being more easily discernible from a distance.</p><p>That said, seabirds may generally need less help than expected. A Vattenfall and Spoor study at its<a href="https://group.vattenfall.com/uk/newsroom/pressreleases/2026/new-aberdeen-study-finds-seabirds-steer-clear-of-turbine-blades" target="_blank" rel="nofollow"> Aberdeen offshore wind farm</a> tracked 2,007 flight paths over 19 months. It found no confirmed collisions, with birds typically keeping 100 to 200 meters away.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/today-i-learned-why-wind-farms-paint-one-turbine-blade-black</link>
                                                                            <description>
                            <![CDATA[ A Norwegian trial cut bird deaths by about 72% with one black blade but a Dutch follow-up experiment found no effect, so newer offshore farms are now trying red. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">V6ts5pdmPRLixEe28DH4fj</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/9pxYTjwg9AimCKJzb6yDgb-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sun, 15 Nov 2026 01:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                <author><![CDATA[ ritoban@nutgraf.agency (Ritoban Mukherjee) ]]></author>                    <dc:creator><![CDATA[ Ritoban Mukherjee ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/cD9joj4H54xYmooW8re3vU-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/9pxYTjwg9AimCKJzb6yDgb-1920-80.jpg">
                                                            <media:credit><![CDATA[Smøla Vindkraftverk]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A black blade experiment at the Smøla wind farm in Central Norway cut bird deaths by 72%]]></media:description>                                                            <media:text><![CDATA[A black blade experiment at the Smøla wind farm in Central Norway cut bird deaths by 72%]]></media:text>
                                <media:title type="plain"><![CDATA[A black blade experiment at the Smøla wind farm in Central Norway cut bird deaths by 72%]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/9pxYTjwg9AimCKJzb6yDgb-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>You'd think that painting a single turbine blade black is a pretty inexpensive fix for accidental bird deaths. It sounds almost too good to be true, which perhaps it is. </p><p>At the Smøla wind farm in Norway, every wind turbine had one of its blade painted black for an extended science experiment back in August 2013. Over the next three years, the number of bird carcasses found fell by 70%. </p><p>However, similar experiments in the Netherlands did not yield positive results, because apparently this fix depends heavily on the local bird species and population size. </p><h2 id="too-many-dead-birds">Too many dead birds</h2><p>There are a total of 68 turbines at the Smøla wind power plant, an archipelago off the coast of Central Norway. Researchers from the Norwegian Institute for Nature Research compared four regular turbines with four neighboring ones that had one blade painted black for this experiment.</p><p>They regularly counted the number of bird carcasses within 100 meters of the wind farm. This continued over a period of more than ten years, combining 7.5 years of baseline data from before painting any turbine blades to 3.5 years of data from after. In 2020, they published their findings in a paper called Ecology and Evolution.</p><p>Turns out that this small change in visual cues reduced the number of bird deaths by over 70%, proving especially effective on the somewhat rare white-tailed sea eagle population of the region. While the researchers noted that this anecdotal data may not hold up to different bird populations, this was still independent real-world confirmation of a separate laboratory experiment that found similar results in 2003.</p><h2 id="how-a-black-blade-is-supposed-to-help">How a black blade is supposed to help</h2><p>What difference does one black blade make? According to researchers, it has to do with how birds process visual cues. </p><p>Animation designers call it motion smearing, where images that move too fast for the human eye to follow are inferred in the brain as one continuous motion. When turbine blades are in motion, birds find it difficult to notice the threat because the unicolor rotating blades look like one giant circle. </p><p>When you paint one of these blades black, it breaks up the pattern so the moving blades are more easily distinguishable to the eye. But as <a href="https://www.audubon.org/magazine/can-painting-wind-turbine-blades-black-really-save-birds" target="_blank" rel="nofollow">Audubon reported</a>,  these are preliminary findings that won't translate to all bird species equally — especially ones that don't have the visual acuity to notice the difference. There's also a significant amount of effort involved, since painting turbines in operation requires both lifts and specialist crews who can navigate the height. </p><h2 id="when-the-trick-stops-working">When the trick stops working</h2><p>Most of the land experiments so far have stuck with black, but at sea, a new trial is experimenting with a different color.</p><p>Dutch consortium Ecowende is fitting<a href="https://ocean-energyresources.com/2026/05/10/red-blades-real-world-test-for-nature-inclusive-offshore-wind/" target="_blank" rel="nofollow"> seven Vestas V236-15.0 MW turbines</a> with one red 115.5-meter blade each. Red was chosen after weighing options against criteria such as durability and heat management, apart from the obvious benefit of being more easily discernible from a distance.</p><p>That said, seabirds may generally need less help than expected. A Vattenfall and Spoor study at its<a href="https://group.vattenfall.com/uk/newsroom/pressreleases/2026/new-aberdeen-study-finds-seabirds-steer-clear-of-turbine-blades" target="_blank" rel="nofollow"> Aberdeen offshore wind farm</a> tracked 2,007 flight paths over 19 months. It found no confirmed collisions, with birds typically keeping 100 to 200 meters away.</p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Google turned to AMD Epyc, not its own Axion chips, to secure its first confidential TPU ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Google's new TPU 8tx pairs TPU 8t silicon with 5th Gen AMD Epyc "Turin" hosts to deliver confidential AI inference</strong></li><li><strong>The AI juggernaut passed over its own Axion CPU, which hosts the 8t and 8i, likely because its audited confidential stack is x86-only for now</strong></li><li><strong>The confidential TPU joins a growing number of chips even as key security details remain unpublished, even as Google and Nvidia already offer confidential GPUs to cloud customers</strong></li></ul><p>Google has added two more systems to its eighth-generation TPU family, and the one built for confidential AI leverages a host processor Google didn't design.</p><p>The TPU 8tx, detailed alongside the dual-chip TPU 8i Max in a <a href="https://cloud.google.com/blog/topics/ai-infrastructure/tpu-8tx-and-tpu-8i-max" target="_blank" rel="nofollow">Google Cloud blog post</a>, pairs Google's accelerator silicon with 5th Gen AMD Epyc CPUs.</p><p>AMD <a href="https://newsroom.amd.com/news/amd-epyc-expands-leadership-foundation-ai/" target="_blank" rel="nofollow">also announced</a> that Epyc forms the host foundation for what both companies call the first cloud TPU system with hardware-enforced confidential computing.</p><h2 id="the-same-chip-with-a-new-trust-boundary-in-tow">The same chip with a new trust boundary in tow?</h2><p>Google's <a href="https://www.techradar.com/pro/google-cloud-unveils-eighth-generation-tpus-built-to-support-an-agentic-era" target="_blank">April 2026 launch</a> split its eighth generation hardware into the TPU 8t for training and the TPU 8i for inference. Google bills this as the first time it has introduced four TPU systems in one year, but as <a href="https://www.servethehome.com/?p=106986" target="_blank" rel="nofollow">ServeTheHome</a> points out, the two newcomers are system configurations built on existing chips rather than new designs. </p><p>As per Taiwanese outlet <a href="https://www.inside.com.tw/article/42841-google-amd-tpu-8tx-8i-max-agentic-ai" target="_blank" rel="nofollow">INSIDE, </a>the 8tx inherits the 8t's 12.6 petaflops of FP4 compute and 216GB of HBM, but swaps in the lower-latency Boardfly network used by the 8i. Google claims up to 2x lower latency than TPU 8t with a near-zero performance penalty, and pitches the system at regulated enterprises and governments that want to fine-tune, distill, and train agents on private data without moving it between clusters.</p><p>AMD says Epyc handles orchestration, data preparation, and the high-throughput link to the accelerators. Its Secure Encrypted Virtualization (SEV) works with the TPU to extend the protected environment out to the accelerators, checking each one's identity and security configuration before granting direct access to protected memory</p><p>Prompts, model weights, and intermediate data stay encrypted in host memory and in transit, according to AMD, with hardware isolation while the TPU runs. Google, for its part, described the result as "keeping sensitive data and model weights protected and isolated even while in active use."</p><h2 id="skipping-the-homegrown-solution">Skipping the homegrown solution?</h2><p>Google hasn't directly said why it passed over Axion, the Neoverse V2-based chip it unveiled in 2024 as its <a href="https://www.techradar.com/pro/google-unveils-its-first-data-center-processor-as-it-joins-microsoft-and-amazon-in-a-billion-dollar-hardware-race-intel-and-amd-will-be-nervous-to-see-strategic-partners-turn-into-formidable-x86-rivals" target="_blank">first data center processor</a>. Its own documentation offers a strong hint, however: Google's <a href="https://docs.cloud.google.com/confidential-computing/confidential-vm/docs/supported-configurations" target="_blank" rel="nofollow">list of supported Confidential VM configurations</a> runs entirely on AMD SEV and Intel TDX machines, with no Axion-based C4A or N4A instances, and states that TPUs can't currently be attached to Confidential VMs at all.</p><p>AMD's release adds that Google security experts have performed deep reviews of AMD's security processor firmware, and that SEV already underpins Google's Confidential VMs, Confidential GKE, Confidential Space, and confidential Spark and Dataflow services.</p><p>It could be argued here that confidential computing is only as credible as the years spent attacking it, and Google's audited chain of trust runs through x86. Building an equivalent on Axion would mean qualifying new hardware before shipping a product whose entire selling point is the "trust" it comes with, making it a moot point.</p><p>The move however is not a win for AMD's latest silicon; ServeTheHome reported that Google chose Turin to match the TPU 8t's PCIe Gen5 x16 interface, noting that a Gen6 host such as <a href="https://www.techradar.com/pro/amds-venice-is-the-first-2nm-x86-server-chip-but-the-stack-doesnt-fully-ship-until-2027" target="_blank">Venice</a> would add no accelerator bandwidth, and tying the decision to Google's 2027 qualification schedule. Venice is AMD's 2nm flagship; the 8tx runs on a chip based on a family of CPUs that <a href="https://www.techradar.com/pro/meet-amds-new-192-core-cpu-monster-the-epyc-9965-is-its-most-expensive-processor-ever-at-nearly-usd15-000" target="_blank">launched in October 2024</a>.</p><p>Mark Lohmeyer, who runs AI and computing infrastructure at Google Cloud, told reporters that select customers will be able to deploy the 8tx in their own data centers, Data Center Knowledge reported, noting that Oracle recently said it will offer TPUs on its cloud. INSIDE quoted Lohmeyer arguing that the technology lowers the risk of physical operators reaching sensitive data, even in third-party facilities.</p><p>As TPUs spread beyond Google's own campuses, hardware attestation becomes the mechanism that lets a model owner place weights on hardware someone else runs. That interestingly puts AMD as a crucial part at the root of Google's off-campus trust offerings, a bigger role than the 8tx's niche positioning would initially suggest.</p><p>It also revives August's rumor: A SemiAnalysis client note reported by <a href="https://www.tomshardware.com/tech-industry/artificial-intelligence/google-reportedly-taps-amd-to-design-next-generation-tpu-hybrid-ai-asic-could-integrate-on-package-cpu-cores-for-reinforcement-learning" target="_blank" rel="nofollow">Tom's Hardware</a>, said "Market chatter suggests [Google] is working with AMD on a TPU project in the v10 generation," possibly with CPU cores on the same package. The 8tx confirms neither claim as its Epyc chips sit in separate sockets, and nothing announced on November 9 involves AMD in accelerator design, but it does lend credence to potential future collaborations between the chip designer and the AI juggernaut.</p><p>The news saw AMD closing at $508.58 on November 9, up 13.26%, after trading as high as $515.18 on more than twice its recent average volume and while not all of it can be attributed to this particular development even as Oppenheimer's Rick Schafer upgraded AMD to Outperform with a $600 target, calling its server and data center success "undeniable", one could argue that it was a sizeable, if not the overwhelming catalyst in play that day.</p><p>The TPU 8tx arguably won't turn AMD's Epyc into a formidable TPU business overnight, and Google has left enough unpublished that its confidential computing claims deserve scrutiny once customers can test them.</p><p>However, one does need to account for the fact that when Google needed a host CPU to stake its first confidential TPU on, it picked AMD over its own chip, and that endorsement will outlast one day's share price increase as the latter aims to broaden its AI business even as its CEO <a href="https://www.techradar.com/computing/gpu/amd-ceo-enrages-gamers-by-saying-its-now-a-data-center-first-company-and-predicts-pc-sales-will-plummet-by-a-double-digit-percentage-in-2027" target="_blank">recently branded the company as "Data center-first" </a>entity even as PC sales continue to plummet.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/google-turned-to-amd-epyc-not-its-own-axion-chips-to-secure-its-first-confidential-tpu</link>
                                                                            <description>
                            <![CDATA[ Google builds its first confidential TPU on AMD Epyc rather than its own Axion chip. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">4uvYw4949V59ejPwgauF99</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/MVfiXV7sRq7dWyFz2yiRHM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sun, 15 Nov 2026 00:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                <author><![CDATA[ Rahimnoorali11@gmail.com (Rahim Amir) ]]></author>                    <dc:creator><![CDATA[ Rahim Amir ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/9xKZFBamtEZKSChRvywbPB-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rahim Amir is a UAE-based tech writer who enjoys building PCs as much as he enjoys writing about them. He has been professionally writing about PC hardware since 2023, focusing on buyer’s guides, hardware reviews, and sponsored content and features related to tech.&lt;br&gt;&lt;br&gt;Having built hundreds of gaming PCs and being an avid gamer in his spare time, Rahim tends to have stronger opinions about hardware than most. This is particularly on display when he gets his way with powerful, but minimalistic RGB builds even as Small Form Factor (SFF) PCs come a close second.&lt;br&gt;&lt;br&gt;In addition to his contributions to TechRadar, Rahim’s work has also been featured on Game Rant and financial news websites.&lt;br&gt;&lt;br&gt;When he’s not working, you can find him playing DotA with friends or schmoozing to take the world over in Civilization. Alternatively, you can find him binging through the entirety of the Lord of The Rings universe with extended editions in play where applicable.&lt;br&gt;&lt;br&gt;You can currently catch Rahim grinding Path of Exile 2, complaining about his (extremely low) unique loot drop rate, or actively participating in one of the numerous (and heated) debates centered around Tolkien&#039;s universe on multiple forums daily.&lt;br&gt;&lt;br&gt;If you have a PC build or a Satisfactory playthrough in progress, he is likely to have some advice to send your way, especially regarding verticality being key for the latter. For the former, Rahim enjoys all aspects of the process including researching the components he will eventually use, benchmarking the latest and greatest hardware he can get his hands on, and somewhat surprisingly, cable management once he gets his latest build to POST.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/MVfiXV7sRq7dWyFz2yiRHM-1920-80.jpg">
                                                            <media:credit><![CDATA[AMD]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A Promotional Rendering Of An AMD Epyc Processor]]></media:description>                                                            <media:text><![CDATA[A Promotional Rendering Of An AMD Epyc Processor]]></media:text>
                                <media:title type="plain"><![CDATA[A Promotional Rendering Of An AMD Epyc Processor]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/MVfiXV7sRq7dWyFz2yiRHM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Google's new TPU 8tx pairs TPU 8t silicon with 5th Gen AMD Epyc "Turin" hosts to deliver confidential AI inference</strong></li><li><strong>The AI juggernaut passed over its own Axion CPU, which hosts the 8t and 8i, likely because its audited confidential stack is x86-only for now</strong></li><li><strong>The confidential TPU joins a growing number of chips even as key security details remain unpublished, even as Google and Nvidia already offer confidential GPUs to cloud customers</strong></li></ul><p>Google has added two more systems to its eighth-generation TPU family, and the one built for confidential AI leverages a host processor Google didn't design.</p><p>The TPU 8tx, detailed alongside the dual-chip TPU 8i Max in a <a href="https://cloud.google.com/blog/topics/ai-infrastructure/tpu-8tx-and-tpu-8i-max" target="_blank" rel="nofollow">Google Cloud blog post</a>, pairs Google's accelerator silicon with 5th Gen AMD Epyc CPUs.</p><p>AMD <a href="https://newsroom.amd.com/news/amd-epyc-expands-leadership-foundation-ai/" target="_blank" rel="nofollow">also announced</a> that Epyc forms the host foundation for what both companies call the first cloud TPU system with hardware-enforced confidential computing.</p><h2 id="the-same-chip-with-a-new-trust-boundary-in-tow">The same chip with a new trust boundary in tow?</h2><p>Google's <a href="https://www.techradar.com/pro/google-cloud-unveils-eighth-generation-tpus-built-to-support-an-agentic-era" target="_blank">April 2026 launch</a> split its eighth generation hardware into the TPU 8t for training and the TPU 8i for inference. Google bills this as the first time it has introduced four TPU systems in one year, but as <a href="https://www.servethehome.com/?p=106986" target="_blank" rel="nofollow">ServeTheHome</a> points out, the two newcomers are system configurations built on existing chips rather than new designs. </p><p>As per Taiwanese outlet <a href="https://www.inside.com.tw/article/42841-google-amd-tpu-8tx-8i-max-agentic-ai" target="_blank" rel="nofollow">INSIDE, </a>the 8tx inherits the 8t's 12.6 petaflops of FP4 compute and 216GB of HBM, but swaps in the lower-latency Boardfly network used by the 8i. Google claims up to 2x lower latency than TPU 8t with a near-zero performance penalty, and pitches the system at regulated enterprises and governments that want to fine-tune, distill, and train agents on private data without moving it between clusters.</p><p>AMD says Epyc handles orchestration, data preparation, and the high-throughput link to the accelerators. Its Secure Encrypted Virtualization (SEV) works with the TPU to extend the protected environment out to the accelerators, checking each one's identity and security configuration before granting direct access to protected memory</p><p>Prompts, model weights, and intermediate data stay encrypted in host memory and in transit, according to AMD, with hardware isolation while the TPU runs. Google, for its part, described the result as "keeping sensitive data and model weights protected and isolated even while in active use."</p><h2 id="skipping-the-homegrown-solution">Skipping the homegrown solution?</h2><p>Google hasn't directly said why it passed over Axion, the Neoverse V2-based chip it unveiled in 2024 as its <a href="https://www.techradar.com/pro/google-unveils-its-first-data-center-processor-as-it-joins-microsoft-and-amazon-in-a-billion-dollar-hardware-race-intel-and-amd-will-be-nervous-to-see-strategic-partners-turn-into-formidable-x86-rivals" target="_blank">first data center processor</a>. Its own documentation offers a strong hint, however: Google's <a href="https://docs.cloud.google.com/confidential-computing/confidential-vm/docs/supported-configurations" target="_blank" rel="nofollow">list of supported Confidential VM configurations</a> runs entirely on AMD SEV and Intel TDX machines, with no Axion-based C4A or N4A instances, and states that TPUs can't currently be attached to Confidential VMs at all.</p><p>AMD's release adds that Google security experts have performed deep reviews of AMD's security processor firmware, and that SEV already underpins Google's Confidential VMs, Confidential GKE, Confidential Space, and confidential Spark and Dataflow services.</p><p>It could be argued here that confidential computing is only as credible as the years spent attacking it, and Google's audited chain of trust runs through x86. Building an equivalent on Axion would mean qualifying new hardware before shipping a product whose entire selling point is the "trust" it comes with, making it a moot point.</p><p>The move however is not a win for AMD's latest silicon; ServeTheHome reported that Google chose Turin to match the TPU 8t's PCIe Gen5 x16 interface, noting that a Gen6 host such as <a href="https://www.techradar.com/pro/amds-venice-is-the-first-2nm-x86-server-chip-but-the-stack-doesnt-fully-ship-until-2027" target="_blank">Venice</a> would add no accelerator bandwidth, and tying the decision to Google's 2027 qualification schedule. Venice is AMD's 2nm flagship; the 8tx runs on a chip based on a family of CPUs that <a href="https://www.techradar.com/pro/meet-amds-new-192-core-cpu-monster-the-epyc-9965-is-its-most-expensive-processor-ever-at-nearly-usd15-000" target="_blank">launched in October 2024</a>.</p><p>Mark Lohmeyer, who runs AI and computing infrastructure at Google Cloud, told reporters that select customers will be able to deploy the 8tx in their own data centers, Data Center Knowledge reported, noting that Oracle recently said it will offer TPUs on its cloud. INSIDE quoted Lohmeyer arguing that the technology lowers the risk of physical operators reaching sensitive data, even in third-party facilities.</p><p>As TPUs spread beyond Google's own campuses, hardware attestation becomes the mechanism that lets a model owner place weights on hardware someone else runs. That interestingly puts AMD as a crucial part at the root of Google's off-campus trust offerings, a bigger role than the 8tx's niche positioning would initially suggest.</p><p>It also revives August's rumor: A SemiAnalysis client note reported by <a href="https://www.tomshardware.com/tech-industry/artificial-intelligence/google-reportedly-taps-amd-to-design-next-generation-tpu-hybrid-ai-asic-could-integrate-on-package-cpu-cores-for-reinforcement-learning" target="_blank" rel="nofollow">Tom's Hardware</a>, said "Market chatter suggests [Google] is working with AMD on a TPU project in the v10 generation," possibly with CPU cores on the same package. The 8tx confirms neither claim as its Epyc chips sit in separate sockets, and nothing announced on November 9 involves AMD in accelerator design, but it does lend credence to potential future collaborations between the chip designer and the AI juggernaut.</p><p>The news saw AMD closing at $508.58 on November 9, up 13.26%, after trading as high as $515.18 on more than twice its recent average volume and while not all of it can be attributed to this particular development even as Oppenheimer's Rick Schafer upgraded AMD to Outperform with a $600 target, calling its server and data center success "undeniable", one could argue that it was a sizeable, if not the overwhelming catalyst in play that day.</p><p>The TPU 8tx arguably won't turn AMD's Epyc into a formidable TPU business overnight, and Google has left enough unpublished that its confidential computing claims deserve scrutiny once customers can test them.</p><p>However, one does need to account for the fact that when Google needed a host CPU to stake its first confidential TPU on, it picked AMD over its own chip, and that endorsement will outlast one day's share price increase as the latter aims to broaden its AI business even as its CEO <a href="https://www.techradar.com/computing/gpu/amd-ceo-enrages-gamers-by-saying-its-now-a-data-center-first-company-and-predicts-pc-sales-will-plummet-by-a-double-digit-percentage-in-2027" target="_blank">recently branded the company as "Data center-first" </a>entity even as PC sales continue to plummet.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Quote of the day by Slack co-founder Stewart Butterfield: 'What we are selling is not the software product… it's organizational transformation" ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Although Slack is now a widely used communications tool deployed by small businesses and enterprises alike, it didn't always start off that way. </p><p>Its creator organization, Tiny Speck, was actually founded as a video games company - but it was the internal system built to manage workflows and communications that gave the company its big break.</p><h2 id="corporate-relief">Corporate relief</h2><p>Slack's co-founder and former CEO, Stewart Butterfield, was describing his vision for the service in an <a href="https://medium.com/@stewart/we-dont-sell-saddles-here-4c59524d650d" target="_blank" rel="nofollow">internal memo</a>, which he later published to a wider audience.</p><div  class="fancy-box"><div class="fancy_box-title">Quote of the day</div><div class="fancy_box_body"><p class="fancy-box__body-text">This article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. <a data-analytics-id="inline-link" href="https://www.techradar.com/tag/qotd">Read the full series here</a>.</p></div></div><p>This was still when the team was known as Tiny Speck. By framing the software in this way, he suggested that what organizations would buy into was a new way of doing things that Slack represented. The result of adopting Slack in the way teams worked would itself be the product, rather than any particular application.</p><p>He clarified that the team was selling a reduction in information overload, relief from stress and the ability to search through corporate archives with ease – a task otherwise laborious and cumbersome.    </p><h2 id="the-modern-workplace">The modern workplace</h2><p>Slack has positioned itself as an essential tool among many others, including Microsoft Teams, in the broader business landscape. </p><p>Even after its massive $28 billion acquisition by Salesforce, it hasn't lost its identity – and has even innovated by adding new features, especially in terms of integrations and custom AI instructions. Using MCP, it can connect with external apps and let members work within other tools. </p><p>It has also added deep research and code execution functions with "big mode" for those working on larger projects. Building on this, the team is planning to deploy AI bots within channels and allow agents to be part of the fabric of communication and organization.  </p><div style="min-height: 250px;">                                <div class="kwizly-quiz kwizly-OdvAJe"></div>                            </div>                            <script src="https://kwizly.com/embed/OdvAJe.js" async=""></script> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/quote-of-the-day-by-slack-co-founder-stewart-butterfield-what-we-are-selling-is-not-the-software-product-what-were-selling-is-organizational-transformation-a-manifesto-for-workplace-evolution</link>
                                                                            <description>
                            <![CDATA[ Slack is one of the leading lights in the move to more collaborative and interactive digital workspaces ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">xgxRmrHqh3mkf3eqiSWPoK</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/q4U9vaV9TdPhLmmqfDt8e9-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 23:00:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Keumars Afifi-Sabet ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/baEeYWYTHEpvddufVqymoA-320-70.jpeg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Keumars Afifi-Sabet is a freelance contributor for Tech Radar and Technology Editor for Live Science. He has written for a variety of publications including ITPro, The Week Digital and ComputerActive. He has worked as a technology journalist for more than five years, having previously held the role of features editor with ITPro. In his previous role, he oversaw the commissioning and publishing of long form in areas including AI, cyber security, cloud computing and digital transformation.&lt;/p&gt;&lt;p&gt;An NCTJ-qualified journalist who specialises in technology, his path into journalism began at university. He immersed himself in student media while studying for a degree in biomedical sciences at Queen Mary, University of London. After graduating, Keumars wrote for a variety of local and national publications as a freelancer, including The Independent, The Observer, and Metro. While studying for his NCTJ certification, his work was commended in the category of ‘Top Scoop’ in the 2017 NCTJ awards. He’s also registered as a foundational chartered manager with the Chartered Management Institute (CMI), having qualified as a Level 3 Team leader with distinction in 2023.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/q4U9vaV9TdPhLmmqfDt8e9-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images/Bloomberg]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[stewart butterfield wearing sunglasses outdoors]]></media:description>                                                            <media:text><![CDATA[stewart butterfield wearing sunglasses outdoors]]></media:text>
                                <media:title type="plain"><![CDATA[stewart butterfield wearing sunglasses outdoors]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/q4U9vaV9TdPhLmmqfDt8e9-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Although Slack is now a widely used communications tool deployed by small businesses and enterprises alike, it didn't always start off that way. </p><p>Its creator organization, Tiny Speck, was actually founded as a video games company - but it was the internal system built to manage workflows and communications that gave the company its big break.</p><h2 id="corporate-relief">Corporate relief</h2><p>Slack's co-founder and former CEO, Stewart Butterfield, was describing his vision for the service in an <a href="https://medium.com/@stewart/we-dont-sell-saddles-here-4c59524d650d" target="_blank" rel="nofollow">internal memo</a>, which he later published to a wider audience.</p><div  class="fancy-box"><div class="fancy_box-title">Quote of the day</div><div class="fancy_box_body"><p class="fancy-box__body-text">This article is part of TechRadar Pro's QOTD project to provide an insight into the minds of the brightest and most recognized figures in the technology industry today and in years gone by. <a data-analytics-id="inline-link" href="https://www.techradar.com/tag/qotd">Read the full series here</a>.</p></div></div><p>This was still when the team was known as Tiny Speck. By framing the software in this way, he suggested that what organizations would buy into was a new way of doing things that Slack represented. The result of adopting Slack in the way teams worked would itself be the product, rather than any particular application.</p><p>He clarified that the team was selling a reduction in information overload, relief from stress and the ability to search through corporate archives with ease – a task otherwise laborious and cumbersome.    </p><h2 id="the-modern-workplace">The modern workplace</h2><p>Slack has positioned itself as an essential tool among many others, including Microsoft Teams, in the broader business landscape. </p><p>Even after its massive $28 billion acquisition by Salesforce, it hasn't lost its identity – and has even innovated by adding new features, especially in terms of integrations and custom AI instructions. Using MCP, it can connect with external apps and let members work within other tools. </p><p>It has also added deep research and code execution functions with "big mode" for those working on larger projects. Building on this, the team is planning to deploy AI bots within channels and allow agents to be part of the fabric of communication and organization.  </p><div style="min-height: 250px;">                                <div class="kwizly-quiz kwizly-OdvAJe"></div>                            </div>                            <script src="https://kwizly.com/embed/OdvAJe.js" async=""></script>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Small nuclear reactors could soon power the next generation of AI data centers — modular reactors could soon power Dow Texas chemical plant ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>X-energy receives approval to construct four small modular nuclear reactors</strong></li><li><strong>Each Xe-100 reactor will generate 80 MW of electricity</strong></li><li><strong>The approval ignores critical outstanding safety concerns</strong></li></ul><p>Artificial intelligence companies are increasingly turning toward nuclear energy as electricity demand rises alongside expanding computing infrastructure.</p><p>X-energy has secured federal permission to construct modular reactors at a chemical manufacturing complex in Texas.</p><p>The approval allows the nuclear startup to bring its technology closer to commercial deployment, although operational reactors remain years away.</p><h2 id="x-energy-secures-approval-for-four-nuclear-reactors">X-energy secures approval for four nuclear reactors</h2><p>The Nuclear Regulatory Commission (NRC) has authorized construction of four Xe-100 reactors at a Dow industrial facility in Texas.</p><p>Each unit is designed to generate 80 megawatts (MW) of electricity alongside another 200 MW of thermal energy - so together, the proposed installation would provide 320 MW of electrical generation, giving the chemical manufacturer an additional source of industrial power.</p><p>X-energy describes the regulatory decision as the first construction authorization for an advanced nuclear installation at an American industrial location.</p><p>However, receiving permission to begin construction does not mean the reactors are ready to supply electricity for <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a>.</p><p>The company expects Dow's installation to enter service sometime during the early 2030s, leaving several years before commercial operation begins.</p><p>The proposed technology uses uranium-bearing fuel pellets, with helium gas circulating through the reactor to carry away generated heat.</p><p>X-energy argues this arrangement offers safety advantages over conventional nuclear power systems, although independent experts have questioned those assurances.</p><p>The Union of Concerned Scientists also criticized the regulatory process, particularly the handling of outstanding safety considerations before construction approval.</p><p>“The NRC only accomplished this feat by pushing consideration of critical safety questions to a later date,” said Edwin Lyman, director of nuclear power safety at the Union of Concerned Scientists.</p><p>He argued that regulators should have resolved those outstanding concerns before authorizing construction at the industrial location.</p><h2 id="technology-companies-are-betting-on-nuclear-electricity">Technology companies are betting on nuclear electricity</h2><p>Demand from artificial intelligence infrastructure operators has given emerging nuclear developers another potential source of commercial interest.</p><p>Large technology companies need substantial electricity supplies for computing facilities, particularly as increasingly powerful AI systems require additional processing capacity.</p><p>Conventional electricity providers face growing pressure to accommodate those requirements, creating opportunities for alternative generation technologies.</p><p>Amazon has become one of X-energy's major financial supporters through its Climate Pledge Fund investment.</p><p>The technology company led a $500 million Series C-1 financing round in October 2024 and separately committed to purchasing 5 gigawatts of nuclear generating capacity, an agreement which gives X-energy a significant potential customer as it works toward developing commercial reactors.</p><p>The startup also attracted substantial investor interest through its initial public offering earlier this year, which generated $1 billion in proceeds.</p><p>However, enthusiasm among investors has moderated since the April listing, suggesting that commercial expectations remain sensitive to development timelines.</p><p>The company still faces the challenge of turning regulatory permission and financial commitments into functioning electricity-generating infrastructure.</p><p>Small modular reactors have attracted attention because their developers expect factory-oriented production and smaller installations to simplify construction.</p><p>Those proposed advantages include shorter development schedules and lower costs, although neither benefit has been conclusively established.</p><p>No small modular reactor currently operates commercially in the United States, leaving X-energy's proposed installation dependent on successful construction and regulatory oversight.</p><p>Via <a href="https://techcrunch.com/2026/11/10/x-energy-gets-green-light-to-build-its-first-small-modular-nuclear-reactors/" target="_blank" rel="nofollow">Techcrunch</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/small-nuclear-reactors-could-soon-power-the-next-generation-of-ai-data-centers</link>
                                                                            <description>
                            <![CDATA[ X-energy secures approval for four nuclear reactors in Texas as developers pursue nuclear electricity for increasingly power-hungry AI infrastructure. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">S8c65znLyDz6fcyTTjywig</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/p2uWFBGHtrHTjrYSDny87M-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 22:35:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Space]]></category>
                                                    <category><![CDATA[Science]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/p2uWFBGHtrHTjrYSDny87M-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A data center with racks of servers and lots of lights glowing]]></media:description>                                                            <media:text><![CDATA[A data center with racks of servers and lots of lights glowing]]></media:text>
                                <media:title type="plain"><![CDATA[A data center with racks of servers and lots of lights glowing]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/p2uWFBGHtrHTjrYSDny87M-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>X-energy receives approval to construct four small modular nuclear reactors</strong></li><li><strong>Each Xe-100 reactor will generate 80 MW of electricity</strong></li><li><strong>The approval ignores critical outstanding safety concerns</strong></li></ul><p>Artificial intelligence companies are increasingly turning toward nuclear energy as electricity demand rises alongside expanding computing infrastructure.</p><p>X-energy has secured federal permission to construct modular reactors at a chemical manufacturing complex in Texas.</p><p>The approval allows the nuclear startup to bring its technology closer to commercial deployment, although operational reactors remain years away.</p><h2 id="x-energy-secures-approval-for-four-nuclear-reactors">X-energy secures approval for four nuclear reactors</h2><p>The Nuclear Regulatory Commission (NRC) has authorized construction of four Xe-100 reactors at a Dow industrial facility in Texas.</p><p>Each unit is designed to generate 80 megawatts (MW) of electricity alongside another 200 MW of thermal energy - so together, the proposed installation would provide 320 MW of electrical generation, giving the chemical manufacturer an additional source of industrial power.</p><p>X-energy describes the regulatory decision as the first construction authorization for an advanced nuclear installation at an American industrial location.</p><p>However, receiving permission to begin construction does not mean the reactors are ready to supply electricity for <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a>.</p><p>The company expects Dow's installation to enter service sometime during the early 2030s, leaving several years before commercial operation begins.</p><p>The proposed technology uses uranium-bearing fuel pellets, with helium gas circulating through the reactor to carry away generated heat.</p><p>X-energy argues this arrangement offers safety advantages over conventional nuclear power systems, although independent experts have questioned those assurances.</p><p>The Union of Concerned Scientists also criticized the regulatory process, particularly the handling of outstanding safety considerations before construction approval.</p><p>“The NRC only accomplished this feat by pushing consideration of critical safety questions to a later date,” said Edwin Lyman, director of nuclear power safety at the Union of Concerned Scientists.</p><p>He argued that regulators should have resolved those outstanding concerns before authorizing construction at the industrial location.</p><h2 id="technology-companies-are-betting-on-nuclear-electricity">Technology companies are betting on nuclear electricity</h2><p>Demand from artificial intelligence infrastructure operators has given emerging nuclear developers another potential source of commercial interest.</p><p>Large technology companies need substantial electricity supplies for computing facilities, particularly as increasingly powerful AI systems require additional processing capacity.</p><p>Conventional electricity providers face growing pressure to accommodate those requirements, creating opportunities for alternative generation technologies.</p><p>Amazon has become one of X-energy's major financial supporters through its Climate Pledge Fund investment.</p><p>The technology company led a $500 million Series C-1 financing round in October 2024 and separately committed to purchasing 5 gigawatts of nuclear generating capacity, an agreement which gives X-energy a significant potential customer as it works toward developing commercial reactors.</p><p>The startup also attracted substantial investor interest through its initial public offering earlier this year, which generated $1 billion in proceeds.</p><p>However, enthusiasm among investors has moderated since the April listing, suggesting that commercial expectations remain sensitive to development timelines.</p><p>The company still faces the challenge of turning regulatory permission and financial commitments into functioning electricity-generating infrastructure.</p><p>Small modular reactors have attracted attention because their developers expect factory-oriented production and smaller installations to simplify construction.</p><p>Those proposed advantages include shorter development schedules and lower costs, although neither benefit has been conclusively established.</p><p>No small modular reactor currently operates commercially in the United States, leaving X-energy's proposed installation dependent on successful construction and regulatory oversight.</p><p>Via <a href="https://techcrunch.com/2026/11/10/x-energy-gets-green-light-to-build-its-first-small-modular-nuclear-reactors/" target="_blank" rel="nofollow">Techcrunch</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Yandex users may have lost their data for good following Ukraine strikes — Russian tech giant confirms 2.84% of customer data is gone forever ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Yandex says it recovered 97.16% of customer data affected by Ukrainian drone strikes</strong></li><li><strong>Equipment damage and missing backups left 2.84% of customer data unrecoverable</strong></li><li><strong>Ukrainian drones targeted Yandex's three supercomputers</strong></li></ul><p>Yandex has recovered 97.16% of customer data affected by Ukrainian drone strikes against its Russian data centers, although some information remains permanently inaccessible.</p><p>The Russian technology giant confirmed that the remaining 2.84% could not be restored due to equipment damage and the absence of backup copies.</p><p>Customers who had activated its backup tools before the incidents retained their information, with no losses reported among those accounts.</p><h2 id="three-day-attacks-hit-at-least-three-yandex-facilities">Three-day attacks hit at least three Yandex facilities</h2><p>The attacks began on October 8 2026, when Ukrainian drones struck a Yandex facility in Sasovo, located within Russia’s Ryazan region.</p><p>This site, operating since 2014, housed two of the company's three supercomputers, Chervonenkis and Lyapunov, with respective peak capacities of 29.42 and 20.03 petaflops.</p><p>Further strikes the next day hit Yandex installation at Grabtsevo industrial park in Russia's Kaluga region, which it describes as its largest computing facility and among Russia’s biggest installations for commercial data processing and storage.</p><p>Another attack followed on October 11, when Ukrainian drones struck a facility in Vladimir housing Galushkin, the company's third supercomputer.</p><p>Galushkin has a maximum performance of 20.64 petaflops, adding another significant computing asset to the facilities affected during the three days.</p><p>The company has not disclosed how the lost customer information was distributed across the affected facilities or how many customers were involved.</p><p>In total, about five <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a> belonging to Yandex, Sberbank, and VK were struck across Russia in October 2026.</p><h2 id="yandex-changes-its-backup-arrangements">Yandex changes its backup arrangements</h2><p>Following the incidents, Yandex announced that backup protection would become active automatically whenever customers create new virtual machines through its cloud platform.</p><p>Customers can still switch off the feature through the management console, meaning the new default does not make backup protection compulsory.</p><p>The company also introduced another availability zone, identified as ru-central1-f, operating from a separate physical location to provide additional service continuity.</p><p>The new zone adds another location for hosting workloads, although the announcement does not establish that it can replace information already lost.</p><p>The incidents have also prompted some Russian businesses to examine hosting arrangements outside the country, particularly in neighboring Kazakhstan.</p><p>Yevgeny Svidersky, vice president of cloud services at ITGlobal.com, told Kommersant that interest in Kazakh facilities had increased.</p><p>Two other industry sources claim that certain companies had already started moving operations to facilities beyond Russia's borders.</p><p>Those sources linked the decisions to concerns about further damage to domestic computing infrastructure and possible interruptions to business services.</p><p>However, there is no mention of the specific companies involved, the volume of data or infrastructure being transferred, or whether the move represents a permanent relocation.</p><p>Via <a href="https://united24media.com/world/ukraines-strikes-leave-russias-yandex-cloud-with-284-of-affected-customer-data-lost-forever-24251" target="_blank" rel="nofollow">United24media</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/yandex-users-may-have-lost-their-data-for-good-following-ukraine-strikes-russian-tech-giant-confirms-2-84-of-customer-data-is-gone-forever</link>
                                                                            <description>
                            <![CDATA[ Yandex recovers 97.16% of customer data after Ukrainian drone strikes, but equipment damage and missing backups left remaining information permanently inaccessible. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">QbP5j3sfyhKxWtzvmGAYRU</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/pjysvtTcT6ffUC32W7xqt6-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 22:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Efosa Udinmwen ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/nwRLdPUNG4rWu4Y6nthHDV-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Efosa has been writing about technology for over 7 years, initially driven by curiosity but now fueled by a strong passion for the field. He holds both a Master&#039;s and a PhD in sciences, which provided him with a solid foundation in analytical thinking. Efosa developed a keen interest in technology policy, specifically exploring the intersection of privacy, security, and politics. His research delves into how technological advancements influence regulatory frameworks and societal norms, particularly concerning data protection and cybersecurity.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/pjysvtTcT6ffUC32W7xqt6-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock / Sashkin]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Data center]]></media:description>                                                            <media:text><![CDATA[Data center]]></media:text>
                                <media:title type="plain"><![CDATA[Data center]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/pjysvtTcT6ffUC32W7xqt6-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Yandex says it recovered 97.16% of customer data affected by Ukrainian drone strikes</strong></li><li><strong>Equipment damage and missing backups left 2.84% of customer data unrecoverable</strong></li><li><strong>Ukrainian drones targeted Yandex's three supercomputers</strong></li></ul><p>Yandex has recovered 97.16% of customer data affected by Ukrainian drone strikes against its Russian data centers, although some information remains permanently inaccessible.</p><p>The Russian technology giant confirmed that the remaining 2.84% could not be restored due to equipment damage and the absence of backup copies.</p><p>Customers who had activated its backup tools before the incidents retained their information, with no losses reported among those accounts.</p><h2 id="three-day-attacks-hit-at-least-three-yandex-facilities">Three-day attacks hit at least three Yandex facilities</h2><p>The attacks began on October 8 2026, when Ukrainian drones struck a Yandex facility in Sasovo, located within Russia’s Ryazan region.</p><p>This site, operating since 2014, housed two of the company's three supercomputers, Chervonenkis and Lyapunov, with respective peak capacities of 29.42 and 20.03 petaflops.</p><p>Further strikes the next day hit Yandex installation at Grabtsevo industrial park in Russia's Kaluga region, which it describes as its largest computing facility and among Russia’s biggest installations for commercial data processing and storage.</p><p>Another attack followed on October 11, when Ukrainian drones struck a facility in Vladimir housing Galushkin, the company's third supercomputer.</p><p>Galushkin has a maximum performance of 20.64 petaflops, adding another significant computing asset to the facilities affected during the three days.</p><p>The company has not disclosed how the lost customer information was distributed across the affected facilities or how many customers were involved.</p><p>In total, about five <a href="https://www.techradar.com/pro/best-data-center-proxies">data centers</a> belonging to Yandex, Sberbank, and VK were struck across Russia in October 2026.</p><h2 id="yandex-changes-its-backup-arrangements">Yandex changes its backup arrangements</h2><p>Following the incidents, Yandex announced that backup protection would become active automatically whenever customers create new virtual machines through its cloud platform.</p><p>Customers can still switch off the feature through the management console, meaning the new default does not make backup protection compulsory.</p><p>The company also introduced another availability zone, identified as ru-central1-f, operating from a separate physical location to provide additional service continuity.</p><p>The new zone adds another location for hosting workloads, although the announcement does not establish that it can replace information already lost.</p><p>The incidents have also prompted some Russian businesses to examine hosting arrangements outside the country, particularly in neighboring Kazakhstan.</p><p>Yevgeny Svidersky, vice president of cloud services at ITGlobal.com, told Kommersant that interest in Kazakh facilities had increased.</p><p>Two other industry sources claim that certain companies had already started moving operations to facilities beyond Russia's borders.</p><p>Those sources linked the decisions to concerns about further damage to domestic computing infrastructure and possible interruptions to business services.</p><p>However, there is no mention of the specific companies involved, the volume of data or infrastructure being transferred, or whether the move represents a permanent relocation.</p><p>Via <a href="https://united24media.com/world/ukraines-strikes-leave-russias-yandex-cloud-with-284-of-affected-customer-data-lost-forever-24251" target="_blank" rel="nofollow">United24media</a></p><figure class="van-image-figure  inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="middle" fullscreen="" width="676" height="213" attribution="" endorsement="" class="inline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ 'It’s time to retire the stereotype that gaming is a waste of time': Logitech is giving workers the day off to play video games — just as GTA VI is set to launch ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Logitech is giving workers a day off to play video games</strong></li><li><strong>Its research found gaming skills can positively affect workplace performance</strong></li><li><strong>Logi Game Day comes conveniently ahead of GTA VI launch on November 19 2026</strong></li></ul><p>Logitech has announced a new worker initiative which gives employees a day off to play video games - just days before the release of <a href="https://www.techradar.com/tag/gta-6" target="_blank">GTA VI</a>.</p><p>The company says the thinking behind 'Logi Game Day' is to formally recognize "the cognitive and professional benefits of gaming" - and that the timing is purely coincidental.</p><p>Employees will be able to take one day off from this month until the same period in 2027 to play a game of their choice.</p><h2 id="it-39-s-game-day">It's Game Day</h2><p>“It’s time to retire the stereotype that gaming is a waste of time,” said Hanneke Faber, Logitech’s chief executive officer. “As a company designing products for work and play, we see every day how engagement in one sphere supports performance in the other.” </p><p>The idea is backed up by official findings - the company’s recent Logitech Ctrl Tomorrow study found 81% of gamers believe that gaming builds real-world confidence, reduces anxiety, and correlates with stronger day-to-day work performance. </p><p>Similarly, another company study found a staggering 98.8% of enterprise ITDM respondents reported gaming skills directly translate to workplace capabilities, including strategic thinking, complex problem-solving, and performing under pressure. </p><p>"Play is not the opposite of work," Elaine Laird, Logitech's chief people officer, added. "When people have dedicated time to log off, decompress and immerse themselves in play, they return with clearer focus, renewed energy and fresh approaches to complex challenges."</p><p>The launch of GTA VI is one of the most hotly-anticipated events in entertainment history, with millions of gamers around the world set to download and get playing on its November 19 release date.</p><p>However that is a working day for many, so worldwide productivity may see somewhat of a drop - with one report suggesting the US economy alone could <a href="https://www.ibtimes.co.uk/gta-6-launch-billion-dollar-sick-day-1818190" target="_blank" rel="nofollow">lose as much as $1 billion in productivity</a> as 1.5 million workers potentially skip work.</p><p>For its part, Logitech will be hosting a streaming event at its Americas headquarters to celebrate Game Day.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/its-time-to-retire-the-stereotype-that-gaming-is-a-waste-of-time-logitech-is-giving-workers-the-day-off-to-play-video-games-just-as-gta-vi-is-set-to-launch</link>
                                                                            <description>
                            <![CDATA[ Logitech research found a strong link between gaming and professional performance. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">nPnEHxw2ayiNf2jMN4ZbmF</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/BqQtezEiLZaYHLpwR3rtsm-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 19:05:00 +0000</pubDate>                                                                                                                                <updated>Mon, 16 Nov 2026 08:34:45 +0000</updated>
                                                                                                                                            <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Gaming]]></category>
                                                                                                                    <dc:creator><![CDATA[ Mike Moore ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/vinm2oPWMvB8yMg7qLhtxg-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Mike Moore is Deputy Editor at TechRadar Pro. He has worked as a B2B and B2C technology journalist for over a decade, including at one of the UK&#039;s leading national newspapers and fellow Future title ITProPortal, covering everything from cybersecurity to phone reviews to VR at the Winter Olympics.&lt;/p&gt;&lt;p&gt;&lt;br&gt;&lt;/p&gt;&lt;p&gt;Mike is the main editorial contact for TechRadar Pro, responsible for the news content across the site, as well as managing the contributed content. PRs looking to pitch news stories, bylines/analysis pieces or event invitations should get in contact via the email address mentioned above.&lt;/p&gt;&lt;p&gt;&lt;br&gt;&lt;/p&gt;&lt;p&gt;He has a Masters degree in American Studies from the University of Nottingham, along with a BA in American &amp; English Studies from the same institution. When he&#039;s not keeping track of all the latest enterprise and workplace trends, he can most likely be found watching, following or taking part in some kind of sport.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/BqQtezEiLZaYHLpwR3rtsm-1920-80.jpg">
                                                            <media:credit><![CDATA[Final]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[A gamer using the VR3000 EX gaming headset to play a game on PlayStation 5.]]></media:description>                                                            <media:text><![CDATA[A gamer using the VR3000 EX gaming headset to play a game on PlayStation 5.]]></media:text>
                                <media:title type="plain"><![CDATA[A gamer using the VR3000 EX gaming headset to play a game on PlayStation 5.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/BqQtezEiLZaYHLpwR3rtsm-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Logitech is giving workers a day off to play video games</strong></li><li><strong>Its research found gaming skills can positively affect workplace performance</strong></li><li><strong>Logi Game Day comes conveniently ahead of GTA VI launch on November 19 2026</strong></li></ul><p>Logitech has announced a new worker initiative which gives employees a day off to play video games - just days before the release of <a href="https://www.techradar.com/tag/gta-6" target="_blank">GTA VI</a>.</p><p>The company says the thinking behind 'Logi Game Day' is to formally recognize "the cognitive and professional benefits of gaming" - and that the timing is purely coincidental.</p><p>Employees will be able to take one day off from this month until the same period in 2027 to play a game of their choice.</p><h2 id="it-39-s-game-day">It's Game Day</h2><p>“It’s time to retire the stereotype that gaming is a waste of time,” said Hanneke Faber, Logitech’s chief executive officer. “As a company designing products for work and play, we see every day how engagement in one sphere supports performance in the other.” </p><p>The idea is backed up by official findings - the company’s recent Logitech Ctrl Tomorrow study found 81% of gamers believe that gaming builds real-world confidence, reduces anxiety, and correlates with stronger day-to-day work performance. </p><p>Similarly, another company study found a staggering 98.8% of enterprise ITDM respondents reported gaming skills directly translate to workplace capabilities, including strategic thinking, complex problem-solving, and performing under pressure. </p><p>"Play is not the opposite of work," Elaine Laird, Logitech's chief people officer, added. "When people have dedicated time to log off, decompress and immerse themselves in play, they return with clearer focus, renewed energy and fresh approaches to complex challenges."</p><p>The launch of GTA VI is one of the most hotly-anticipated events in entertainment history, with millions of gamers around the world set to download and get playing on its November 19 release date.</p><p>However that is a working day for many, so worldwide productivity may see somewhat of a drop - with one report suggesting the US economy alone could <a href="https://www.ibtimes.co.uk/gta-6-launch-billion-dollar-sick-day-1818190" target="_blank" rel="nofollow">lose as much as $1 billion in productivity</a> as 1.5 million workers potentially skip work.</p><p>For its part, Logitech will be hosting a streaming event at its Americas headquarters to celebrate Game Day.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ TikTok Head of Global Security on how SMBs can boost their cybersecurity for free ]]></title>
                                                                                                <dc:content><![CDATA[ <p>Nearly<a href="https://ec.europa.eu/eurostat/statistics-explained/index.php/Social_media_-_statistics_on_the_use_by_enterprises" target="_blank" rel="nofollow"> two-thirds</a> of businesses use social media, and for many smaller firms it’s where customers discover, get to know, and buy from them. This means that for a small or medium-sized business, a social media account is often its shopfront, sales channel, and customer service desk all at once.</p><p>It pays off. TikTok helped UK small and medium-sized businesses generate<a href="https://tiktokuk.publicfirst.co/report.html" target="_blank" rel="nofollow"> an estimated £3.4bn</a> in additional output in 2025.</p><p>And while a digital-first world can be great for business, scammers and cybercriminals are evolving their tactics to match. In the UK alone, nearly<a href="https://www.gov.uk/government/statistics/cyber-security-breaches-survey-20252026/cyber-security-breaches-survey-20252026#fn:1" target="_blank" rel="nofollow"> half of small businesses</a> have reported being targeted by a cyber breach or attack.</p><p>At TikTok,<a href="https://www.tiktok.com/safety/en/transparency/cg-report" target="_blank" rel="nofollow"> more than 99%</a> of the frauds and scams content we remove is taken down before anyone reports it. But there are plenty of steps businesses should take, and the good news is that it doesn’t require a dedicated tech team or a big budget.</p><h2 id="1-treat-unexpected-offers-as-unverified">1. Treat unexpected offers as unverified</h2><p>The most effective scams don't look like scams. They look like opportunities: an investor who gets in touch out of the blue, a brand that slides into your DMs with a collaboration but needs an upfront fee to "secure the deal", a service promising to double your followers by Friday. They work because they speak to what every founder wants most: funding and growth.</p><p>So treat every unexpected offer as unverified until you've checked it yourself, through the organisation's official website or a contact you already trust, never the link or number in the message. And ignore anyone selling likes, followers or verification. At best, you'll get hollow numbers. At worst, you'll hand over the keys to your account.</p><h2 id="2-lock-down-your-logins">2. Lock down your logins</h2><p>Your login is the front door to the business you've built. Use a unique password of at least 12 characters for every account your business uses, so one leaked password can't unlock the rest. Turn on 2-Step Verification, so anyone who steals your password still can't get in without a second check.</p><p>Where your devices support them, set up passkeys. They replace passwords with something tied to your device, like a fingerprint or face scan, which makes them far harder to steal. Finally, never send a password by email or direct message, however urgent the request or familiar the sender.</p><h2 id="3-know-who-has-access">3. Know who has access</h2><p>In small teams, it's tempting to share one login across everyone: the founder, the part-time social media manager, the freelancer who helped with the launch. It's convenient, but you can't tell who did what, and you can't remove one person without disrupting everyone else.</p><p>Instead, give each team member their own access, and use a password manager to keep credentials organised. Keep your recovery email and phone number up to date, and check security alerts and active devices regularly.</p><p>Most importantly, remove employees, freelancers and agencies as soon as their role ends. Offboarding is one of the most overlooked moments in business security, and one of the easiest to get right.</p><h2 id="4-watch-what-39-s-in-shot">4. Watch what's in shot</h2><p>Behind-the-scenes content, whether that's packing orders or showing how a product is made, is one of the best ways to build a following. It can also give away more than you intend. Before you post, check the background. A delivery label, an order slip or a customer's details left on a packing table can tell a scammer exactly what they need to target you, or the people who buy from you.</p><h2 id="5-report-anything-suspicious">5. Report anything suspicious</h2><p>If something feels off, whether it's an email or something you see on social media, report it. Every report helps platforms spot patterns and act faster, protecting not just your business but others being targeted in a similar way.</p><h2 id="small-habits-serious-protection">Small habits, serious protection</h2><p>None of these steps requires technical expertise or a dedicated tech team. A cyber incident can be costly, but basic protections cost nothing. Whether you’re a one-person operation or a growing team, these habits will help keep the revenue coming in, and the cybercriminals out.</p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/a-cyber-incident-can-be-costly-but-basic-protections-cost-nothing-tiktoks-head-of-global-security-on-how-smbs-can-boost-their-cybersecurity-for-free</link>
                                                                            <description>
                            <![CDATA[ How your SMB can boost cybersecurity. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">gkaskkTyVziBw86WX6Gfzh</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/2iH4kSL6LCJbxYBpQT5iiM-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 14:00:00 +0000</pubDate>                                                                                                                                <updated>Mon, 16 Nov 2026 08:34:45 +0000</updated>
                                                                                                                                            <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Kim Albarella ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/2iH4kSL6LCJbxYBpQT5iiM-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[smb]]></media:description>                                                            <media:text><![CDATA[smb]]></media:text>
                                <media:title type="plain"><![CDATA[smb]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/2iH4kSL6LCJbxYBpQT5iiM-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>Nearly<a href="https://ec.europa.eu/eurostat/statistics-explained/index.php/Social_media_-_statistics_on_the_use_by_enterprises" target="_blank" rel="nofollow"> two-thirds</a> of businesses use social media, and for many smaller firms it’s where customers discover, get to know, and buy from them. This means that for a small or medium-sized business, a social media account is often its shopfront, sales channel, and customer service desk all at once.</p><p>It pays off. TikTok helped UK small and medium-sized businesses generate<a href="https://tiktokuk.publicfirst.co/report.html" target="_blank" rel="nofollow"> an estimated £3.4bn</a> in additional output in 2025.</p><p>And while a digital-first world can be great for business, scammers and cybercriminals are evolving their tactics to match. In the UK alone, nearly<a href="https://www.gov.uk/government/statistics/cyber-security-breaches-survey-20252026/cyber-security-breaches-survey-20252026#fn:1" target="_blank" rel="nofollow"> half of small businesses</a> have reported being targeted by a cyber breach or attack.</p><p>At TikTok,<a href="https://www.tiktok.com/safety/en/transparency/cg-report" target="_blank" rel="nofollow"> more than 99%</a> of the frauds and scams content we remove is taken down before anyone reports it. But there are plenty of steps businesses should take, and the good news is that it doesn’t require a dedicated tech team or a big budget.</p><h2 id="1-treat-unexpected-offers-as-unverified">1. Treat unexpected offers as unverified</h2><p>The most effective scams don't look like scams. They look like opportunities: an investor who gets in touch out of the blue, a brand that slides into your DMs with a collaboration but needs an upfront fee to "secure the deal", a service promising to double your followers by Friday. They work because they speak to what every founder wants most: funding and growth.</p><p>So treat every unexpected offer as unverified until you've checked it yourself, through the organisation's official website or a contact you already trust, never the link or number in the message. And ignore anyone selling likes, followers or verification. At best, you'll get hollow numbers. At worst, you'll hand over the keys to your account.</p><h2 id="2-lock-down-your-logins">2. Lock down your logins</h2><p>Your login is the front door to the business you've built. Use a unique password of at least 12 characters for every account your business uses, so one leaked password can't unlock the rest. Turn on 2-Step Verification, so anyone who steals your password still can't get in without a second check.</p><p>Where your devices support them, set up passkeys. They replace passwords with something tied to your device, like a fingerprint or face scan, which makes them far harder to steal. Finally, never send a password by email or direct message, however urgent the request or familiar the sender.</p><h2 id="3-know-who-has-access">3. Know who has access</h2><p>In small teams, it's tempting to share one login across everyone: the founder, the part-time social media manager, the freelancer who helped with the launch. It's convenient, but you can't tell who did what, and you can't remove one person without disrupting everyone else.</p><p>Instead, give each team member their own access, and use a password manager to keep credentials organised. Keep your recovery email and phone number up to date, and check security alerts and active devices regularly.</p><p>Most importantly, remove employees, freelancers and agencies as soon as their role ends. Offboarding is one of the most overlooked moments in business security, and one of the easiest to get right.</p><h2 id="4-watch-what-39-s-in-shot">4. Watch what's in shot</h2><p>Behind-the-scenes content, whether that's packing orders or showing how a product is made, is one of the best ways to build a following. It can also give away more than you intend. Before you post, check the background. A delivery label, an order slip or a customer's details left on a packing table can tell a scammer exactly what they need to target you, or the people who buy from you.</p><h2 id="5-report-anything-suspicious">5. Report anything suspicious</h2><p>If something feels off, whether it's an email or something you see on social media, report it. Every report helps platforms spot patterns and act faster, protecting not just your business but others being targeted in a similar way.</p><h2 id="small-habits-serious-protection">Small habits, serious protection</h2><p>None of these steps requires technical expertise or a dedicated tech team. A cyber incident can be costly, but basic protections cost nothing. Whether you’re a one-person operation or a growing team, these habits will help keep the revenue coming in, and the cybercriminals out.</p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Gigabyte unveils its Nvidia Spark-powered AI workstation — AI TOP AERO combines Blackwell architecture GPU and up to 128 GB memory, if you can get one ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Gigabyte's AI TOP AERO fits Nvidia's 20-core RTX Spark N1X and up to 128GB of memory into a one-liter Windows 11 desktop</strong></li><li><strong>Three days after its launch date, there is still no price, no regional configuration, and no product listing</strong></li><li><strong>Rival 128GB RTX Spark and DGX Spark boxes cost $5,999 to $7,999, and MSI's top model is already out of stock</strong></li></ul><p>Gigabyte has <a href="https://www.prnewswire.com/news-releases/gigabyte-expands-its-local-ai-portfolio-with-the-launch-of-ai-top-aero-powered-by-nvidia-rtx-spark-302934919.html" target="_blank" rel="nofollow">announced the AI TOP AERO</a>, a roughly one-liter desktop built on Nvidia's RTX Spark platform that pairs a 20-core Arm-based Grace CPU with a Blackwell architecture RTX GPU, up to 128GB of unified LPDDR5X memory and up to a 4TB PCIe Gen5 SSD.</p><p>The company is pitching the machine as a Personal AI Studio for local AI development, content creation, and gaming on Windows 11.</p><p>However the announcement still lacks a price, a list of configurations, or anywhere to place an order, even as its closing line tells buyers to "refer to local GIGABYTE announcements for regional product configurations, pricing and availability"; As of November 13, Gigabyte's own <a href="https://www.gigabyte.com/AI-TOP-PC">AI TOP PC listing</a> showed only the AI TOP 100 B850 and the AI TOP ATOM.</p><h2 id="two-different-kinds-of-sparks-on-one-shelf">Two different kinds of Sparks on one shelf</h2><p>Gigabyte now sells two kinds of Nvidia's Spark platform-based hardware. The AI TOP ATOM is a DGX Spark system built on Nvidia's GB10 chip. It runs the Linux-based DGX OS and, according to Gigabyte's <a href="https://www.gigabyte.com/press/news/2461" target="_blank" rel="nofollow">October 2026 announcement of a 64GB version</a>, can be clustered up to four units deep over built-in ConnectX-7 networking.</p><p>The AERO is an RTX Spark machine built on the N1X Superchip. It runs Windows 11 and lists USB Type-C, 10GbE, and Wi-Fi 7, with no mention of clustering hardware.</p><p>Under the lid, however, the two are close relatives: Nvidia's <a href="https://www.nvidia.com/en-us/products/rtx-spark/specifications/" target="_blank" rel="nofollow">specification table</a> lists the desktop N1X with 6,144 CUDA cores, 20 CPU cores, and a 140W TDP, and <em>The Register</em> describes RTX Spark systems as built on <a href="https://www.theregister.com/systems/2026/10/02/nvidia-debuts-4999-dgx-spark-with-half-the-ram-and-storage-amid-memory-crunch/5300622" target="_blank" rel="nofollow">the same GB10 silicon</a> as DGX Spark. Gigabyte's release, however, states none of those figures, nor the AERO's dimensions, weight, or port count.</p><p>The argument for the AERO is that Windows makes this hardware useful to more people. Jensen Huang, then Nvidia's chief executive, set that tone <a href="https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark" target="_blank" rel="nofollow">at Computex in May 2026</a>, declaring, "This is the new PC. The personal AI computer," in a release naming Acer and Gigabyte as vendors that would follow the first six.</p><p>Gigabyte's version of the pitch leans on LM Studio, Ollama, ComfyUI and Nvidia's Nemotron models, plus <a href="https://blogs.nvidia.com/blog/local-ai-ifa-next-gen-agents-nv-pair-rtx-spark/" target="_blank" rel="nofollow">Nvidia PAIR</a>, a free, open-source router introduced in beta in September that spreads independent inference requests across the PCs on a local network.</p><h2 id="a-launch-date-with-no-availability-and-no-price">A launch date with no availability and no price?</h2><p>Finding a unit available online is just as hard as finding the price tag for an AI workhorse that already finds most of its competition sold out. Gigabyte's silence on cost comes at a time when memory sets the price. Nvidia has raised its own 128GB DGX Spark to $6,950, almost three-quarters more than it cost at launch, and <em>The Register</em> judged at the time that the pricing "doesn't bode well" for RTX Spark notebooks and mini PCs.</p><p>Gigabyte's 64GB AI TOP ATOM, the nearest reference point in its own range, starts at $4,999 with a 2TB SSD in tow. The first RTX Spark desktops to reach checkout are no cheaper. MSI's 1.3-liter EdgeMesa N AI+ <a href="https://www.notebookcheck.net/MSI-releases-new-compact-desktop-PC-with-Nvidia-RTX-Spark-and-6-144-CUDA-cores.1447829.0.html" target="_blank" rel="nofollow">went on sale in the US this week</a> at $4,699 with 64GB and 1TB, and $7,999 with 128GB and 4TB, the latter out of stock when Notebookcheck checked on November 11.</p><p>That is a $3,300 step for the extra memory and storage for a platform that is already seemingly prohibitively expensive for end users. Microsoft lists its 128GB Surface RTX Spark Dev Box at $5,999 with shipping due this month, and the 128GB Surface Laptop Ultra has had lead times stretch to late December within days of its launch.</p><p>Buyers who value capacity over CUDA have another route. GMKtec's EVO-X5 Pro, built on AMD's Gorgon Halo platform, <a href="https://www.techradar.com/pro/usd6-399-thats-how-much-gmktec-sells-its-evo-x5-pro-mini-pc-with-an-amd-max-pro-495-and-192gb-ram-and-thats-the-cheapest-fully-loaded-gorgon-halo-device" target="_blank">launched from $6,399</a> with 192GB of memory, up to 160GB of which can be assigned to graphics. None of the price gouging is likely to ease soon, with one forecast predicting the memory crisis will last <a href="https://www.techradar.com/computing/memory/ram-to-see-meaningful-price-increases-before-2026-is-out-as-the-memory-crisis-is-predicted-to-last-at-least-another-3-years" target="_blank">at least another three years</a>.</p><p>Until Gigabyte's regional offices publish prices and stock, the AI TOP AERO is a spec sheet with a date on it. The hardware is credible, largely because it already ships under other names. The open questions are how well a one-liter chassis cools 140W, how much of that 128GB a Windows workload can really use, and whether the top configuration can be bought at all before the year is out.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/gigabyte-unveils-its-nvidia-spark-powered-ai-workstation-ai-top-aero-combines-blackwell-architecture-gpu-and-up-to-128-gb-memory-if-you-can-get-one</link>
                                                                            <description>
                            <![CDATA[ Three days after its stated launch, the one-liter AI TOP AERO still has no price, no regional configurations and no product page. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">BcwnsRxtoCGnYgLo5oDGCK</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/oN5HTKWqSHhSGqUHoEHizi-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Sat, 14 Nov 2026 01:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                <author><![CDATA[ Rahimnoorali11@gmail.com (Rahim Amir) ]]></author>                    <dc:creator><![CDATA[ Rahim Amir ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/9xKZFBamtEZKSChRvywbPB-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Rahim Amir is a UAE-based tech writer who enjoys building PCs as much as he enjoys writing about them. He has been professionally writing about PC hardware since 2023, focusing on buyer’s guides, hardware reviews, and sponsored content and features related to tech.&lt;br&gt;&lt;br&gt;Having built hundreds of gaming PCs and being an avid gamer in his spare time, Rahim tends to have stronger opinions about hardware than most. This is particularly on display when he gets his way with powerful, but minimalistic RGB builds even as Small Form Factor (SFF) PCs come a close second.&lt;br&gt;&lt;br&gt;In addition to his contributions to TechRadar, Rahim’s work has also been featured on Game Rant and financial news websites.&lt;br&gt;&lt;br&gt;When he’s not working, you can find him playing DotA with friends or schmoozing to take the world over in Civilization. Alternatively, you can find him binging through the entirety of the Lord of The Rings universe with extended editions in play where applicable.&lt;br&gt;&lt;br&gt;You can currently catch Rahim grinding Path of Exile 2, complaining about his (extremely low) unique loot drop rate, or actively participating in one of the numerous (and heated) debates centered around Tolkien&#039;s universe on multiple forums daily.&lt;br&gt;&lt;br&gt;If you have a PC build or a Satisfactory playthrough in progress, he is likely to have some advice to send your way, especially regarding verticality being key for the latter. For the former, Rahim enjoys all aspects of the process including researching the components he will eventually use, benchmarking the latest and greatest hardware he can get his hands on, and somewhat surprisingly, cable management once he gets his latest build to POST.&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/oN5HTKWqSHhSGqUHoEHizi-1920-80.jpg">
                                                            <media:credit><![CDATA[Gigabyte]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Gigabyte AI TOP AERO PC]]></media:description>                                                            <media:text><![CDATA[Gigabyte AI TOP AERO PC]]></media:text>
                                <media:title type="plain"><![CDATA[Gigabyte AI TOP AERO PC]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/oN5HTKWqSHhSGqUHoEHizi-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Gigabyte's AI TOP AERO fits Nvidia's 20-core RTX Spark N1X and up to 128GB of memory into a one-liter Windows 11 desktop</strong></li><li><strong>Three days after its launch date, there is still no price, no regional configuration, and no product listing</strong></li><li><strong>Rival 128GB RTX Spark and DGX Spark boxes cost $5,999 to $7,999, and MSI's top model is already out of stock</strong></li></ul><p>Gigabyte has <a href="https://www.prnewswire.com/news-releases/gigabyte-expands-its-local-ai-portfolio-with-the-launch-of-ai-top-aero-powered-by-nvidia-rtx-spark-302934919.html" target="_blank" rel="nofollow">announced the AI TOP AERO</a>, a roughly one-liter desktop built on Nvidia's RTX Spark platform that pairs a 20-core Arm-based Grace CPU with a Blackwell architecture RTX GPU, up to 128GB of unified LPDDR5X memory and up to a 4TB PCIe Gen5 SSD.</p><p>The company is pitching the machine as a Personal AI Studio for local AI development, content creation, and gaming on Windows 11.</p><p>However the announcement still lacks a price, a list of configurations, or anywhere to place an order, even as its closing line tells buyers to "refer to local GIGABYTE announcements for regional product configurations, pricing and availability"; As of November 13, Gigabyte's own <a href="https://www.gigabyte.com/AI-TOP-PC">AI TOP PC listing</a> showed only the AI TOP 100 B850 and the AI TOP ATOM.</p><h2 id="two-different-kinds-of-sparks-on-one-shelf">Two different kinds of Sparks on one shelf</h2><p>Gigabyte now sells two kinds of Nvidia's Spark platform-based hardware. The AI TOP ATOM is a DGX Spark system built on Nvidia's GB10 chip. It runs the Linux-based DGX OS and, according to Gigabyte's <a href="https://www.gigabyte.com/press/news/2461" target="_blank" rel="nofollow">October 2026 announcement of a 64GB version</a>, can be clustered up to four units deep over built-in ConnectX-7 networking.</p><p>The AERO is an RTX Spark machine built on the N1X Superchip. It runs Windows 11 and lists USB Type-C, 10GbE, and Wi-Fi 7, with no mention of clustering hardware.</p><p>Under the lid, however, the two are close relatives: Nvidia's <a href="https://www.nvidia.com/en-us/products/rtx-spark/specifications/" target="_blank" rel="nofollow">specification table</a> lists the desktop N1X with 6,144 CUDA cores, 20 CPU cores, and a 140W TDP, and <em>The Register</em> describes RTX Spark systems as built on <a href="https://www.theregister.com/systems/2026/10/02/nvidia-debuts-4999-dgx-spark-with-half-the-ram-and-storage-amid-memory-crunch/5300622" target="_blank" rel="nofollow">the same GB10 silicon</a> as DGX Spark. Gigabyte's release, however, states none of those figures, nor the AERO's dimensions, weight, or port count.</p><p>The argument for the AERO is that Windows makes this hardware useful to more people. Jensen Huang, then Nvidia's chief executive, set that tone <a href="https://nvidianews.nvidia.com/news/nvidia-microsoft-windows-pcs-agents-rtx-spark" target="_blank" rel="nofollow">at Computex in May 2026</a>, declaring, "This is the new PC. The personal AI computer," in a release naming Acer and Gigabyte as vendors that would follow the first six.</p><p>Gigabyte's version of the pitch leans on LM Studio, Ollama, ComfyUI and Nvidia's Nemotron models, plus <a href="https://blogs.nvidia.com/blog/local-ai-ifa-next-gen-agents-nv-pair-rtx-spark/" target="_blank" rel="nofollow">Nvidia PAIR</a>, a free, open-source router introduced in beta in September that spreads independent inference requests across the PCs on a local network.</p><h2 id="a-launch-date-with-no-availability-and-no-price">A launch date with no availability and no price?</h2><p>Finding a unit available online is just as hard as finding the price tag for an AI workhorse that already finds most of its competition sold out. Gigabyte's silence on cost comes at a time when memory sets the price. Nvidia has raised its own 128GB DGX Spark to $6,950, almost three-quarters more than it cost at launch, and <em>The Register</em> judged at the time that the pricing "doesn't bode well" for RTX Spark notebooks and mini PCs.</p><p>Gigabyte's 64GB AI TOP ATOM, the nearest reference point in its own range, starts at $4,999 with a 2TB SSD in tow. The first RTX Spark desktops to reach checkout are no cheaper. MSI's 1.3-liter EdgeMesa N AI+ <a href="https://www.notebookcheck.net/MSI-releases-new-compact-desktop-PC-with-Nvidia-RTX-Spark-and-6-144-CUDA-cores.1447829.0.html" target="_blank" rel="nofollow">went on sale in the US this week</a> at $4,699 with 64GB and 1TB, and $7,999 with 128GB and 4TB, the latter out of stock when Notebookcheck checked on November 11.</p><p>That is a $3,300 step for the extra memory and storage for a platform that is already seemingly prohibitively expensive for end users. Microsoft lists its 128GB Surface RTX Spark Dev Box at $5,999 with shipping due this month, and the 128GB Surface Laptop Ultra has had lead times stretch to late December within days of its launch.</p><p>Buyers who value capacity over CUDA have another route. GMKtec's EVO-X5 Pro, built on AMD's Gorgon Halo platform, <a href="https://www.techradar.com/pro/usd6-399-thats-how-much-gmktec-sells-its-evo-x5-pro-mini-pc-with-an-amd-max-pro-495-and-192gb-ram-and-thats-the-cheapest-fully-loaded-gorgon-halo-device" target="_blank">launched from $6,399</a> with 192GB of memory, up to 160GB of which can be assigned to graphics. None of the price gouging is likely to ease soon, with one forecast predicting the memory crisis will last <a href="https://www.techradar.com/computing/memory/ram-to-see-meaningful-price-increases-before-2026-is-out-as-the-memory-crisis-is-predicted-to-last-at-least-another-3-years" target="_blank">at least another three years</a>.</p><p>Until Gigabyte's regional offices publish prices and stock, the AI TOP AERO is a spec sheet with a date on it. The hardware is credible, largely because it already ships under other names. The open questions are how well a one-liter chassis cools 140W, how much of that 128GB a Windows workload can really use, and whether the top configuration can be bought at all before the year is out.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Younger workers are hiding AI usage from bosses as they try to hold on to jobs ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Shadow AI is growing in the workplace, with 29% of Gen Z workers admitting to using it</strong></li><li><strong>26% of Gen Z workers feel their colleagues will view secret AI use negatively</strong></li><li><strong>Meanwhile, 35% of the same group feel dishonest for using AI</strong></li></ul><p>The use of AI in the workplace is increasing – whether it has been officially mandated or otherwise – and some groups of employees are finding it increasingly challenging. </p><p>Research into AI use by Imperial College London’s new Imperial Lifelong Learning division has found that over one third of the Gen Z cohort (35%) feel dishonest for using AI. Additionally, over a quarter (26%) feel that their colleagues or managers will be disappointed. </p><p>The research also shows that while Millennial workers are the group mostly using AI in their day-to-day work, they’re similarly worried about their colleagues’ opinions on AI use. The report surfaces existing concerns over the growth of what has been dubbed “shadow AI” – the use of artificial intelligence chatbots tat has not been approved by the employer.</p><h2 id="complex-relationship">Complex relationship</h2><p>Imperial Lifelong Learning’s research highlights what it calls a “complex” relationship between Gen Z and AI. As well as saying using AI feels dishonest and fearing colleague opinions over its use, Gen Z also engages in the dishonesty, with 29% using AI without their employer’s knowledge – 11% more than the Millennial cohort and over double the figure for Gen X.</p><p>The research covered the use of AI across five industries: engineering, financial services, technology, energy, and healthcare. Analyzing the results, it has also found a sizeable amount – 59% of workers – who believe that there is a skills gap in the industry they work in. </p><h2 id="shadow-ai-rises">Shadow AI rises</h2><p>Underpinning all of this are the concerns over shadow AI. Despite LLM-backed chatbots being widely available since 2022 and gaining increasing power and depth with each model, some organizations still don’t have active AI policies.</p><p>“Our findings point to trepidation amongst UK workers around the use of AI and how it will impact the economy," noted Heather Haseley, Chief Lifelong Learning Officer at Imperial Lifelong Learning.</p><p>"Education has a critical role to play in making sure people - of every generation - can do their jobs fully while using these tools responsibly and transparently, and feel confident in how they are governed. For younger employees, support must start before they enter the workplace. ”</p><p>Where organizations and businesses don’t provide that education and guidance, doubt grows. “Industry leaders we spoke to as part of our research highlighted a disconnect between attitudes towards AI in the workplace and in education,” Haseley added, “where young learners may be penalised for using AI, creating an uncertainty that follows them into work.”</p><p>While Imperial Lifelong Learning cites higher education as one way to “bridge that gap,” it is also something that needs addressing on day one of every new job.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/younger-workers-are-hiding-ai-usage-from-bosses-as-they-try-to-hold-on-to-jobs</link>
                                                                            <description>
                            <![CDATA[ Gen Z workers using AI for their role increasingly feel that they have to hide their actions, fearing negative reactions from colleagues and managers. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">tMJrMfm6WXXWryCrNjT4zA</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/CUB5dHBZDJoczb2xqmTYZD-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 19:05:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                                                                                    <dc:creator><![CDATA[ Christian Cawley ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/zBDYnjPnB2XPvhKbYX9Kuc-320-70.png ]]></dc:source>
                                                                <dc:description><![CDATA[ &lt;p&gt;Christian Cawley has extensive experience as a writer and editor in consumer electronics, IT and entertainment media. He has contributed to TechRadar since 2017 and has been published in Computer Weekly, Linux Format, ComputerActive, and other publications. &lt;/p&gt;&lt;p&gt;Beyond TechRadar, he heads up the team at smart home website Matter Alpha, and writes about retro gaming at Gaming Retro. &lt;/p&gt;&lt;p&gt;Formerly the editor responsible for Linux, Security, Programming, and DIY at MakeUseOf, Christian previously worked as a desktop and software support specialist in the public and private sectors.&lt;br&gt;&lt;/p&gt; ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/CUB5dHBZDJoczb2xqmTYZD-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images / PeopleImages]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Laptop, headache and man in cafe.]]></media:description>                                                            <media:text><![CDATA[Laptop, headache and man in cafe.]]></media:text>
                                <media:title type="plain"><![CDATA[Laptop, headache and man in cafe.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/CUB5dHBZDJoczb2xqmTYZD-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Shadow AI is growing in the workplace, with 29% of Gen Z workers admitting to using it</strong></li><li><strong>26% of Gen Z workers feel their colleagues will view secret AI use negatively</strong></li><li><strong>Meanwhile, 35% of the same group feel dishonest for using AI</strong></li></ul><p>The use of AI in the workplace is increasing – whether it has been officially mandated or otherwise – and some groups of employees are finding it increasingly challenging. </p><p>Research into AI use by Imperial College London’s new Imperial Lifelong Learning division has found that over one third of the Gen Z cohort (35%) feel dishonest for using AI. Additionally, over a quarter (26%) feel that their colleagues or managers will be disappointed. </p><p>The research also shows that while Millennial workers are the group mostly using AI in their day-to-day work, they’re similarly worried about their colleagues’ opinions on AI use. The report surfaces existing concerns over the growth of what has been dubbed “shadow AI” – the use of artificial intelligence chatbots tat has not been approved by the employer.</p><h2 id="complex-relationship">Complex relationship</h2><p>Imperial Lifelong Learning’s research highlights what it calls a “complex” relationship between Gen Z and AI. As well as saying using AI feels dishonest and fearing colleague opinions over its use, Gen Z also engages in the dishonesty, with 29% using AI without their employer’s knowledge – 11% more than the Millennial cohort and over double the figure for Gen X.</p><p>The research covered the use of AI across five industries: engineering, financial services, technology, energy, and healthcare. Analyzing the results, it has also found a sizeable amount – 59% of workers – who believe that there is a skills gap in the industry they work in. </p><h2 id="shadow-ai-rises">Shadow AI rises</h2><p>Underpinning all of this are the concerns over shadow AI. Despite LLM-backed chatbots being widely available since 2022 and gaining increasing power and depth with each model, some organizations still don’t have active AI policies.</p><p>“Our findings point to trepidation amongst UK workers around the use of AI and how it will impact the economy," noted Heather Haseley, Chief Lifelong Learning Officer at Imperial Lifelong Learning.</p><p>"Education has a critical role to play in making sure people - of every generation - can do their jobs fully while using these tools responsibly and transparently, and feel confident in how they are governed. For younger employees, support must start before they enter the workplace. ”</p><p>Where organizations and businesses don’t provide that education and guidance, doubt grows. “Industry leaders we spoke to as part of our research highlighted a disconnect between attitudes towards AI in the workplace and in education,” Haseley added, “where young learners may be penalised for using AI, creating an uncertainty that follows them into work.”</p><p>While Imperial Lifelong Learning cites higher education as one way to “bridge that gap,” it is also something that needs addressing on day one of every new job.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ FBI seizes tools used by North Korean hackers, freezes $1.2bilion in stolen Binance funds ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>The us DoJ said it obtained a court order to "freeze or seize" $1.2 billion out of the $6.7bn stolen from Binance</strong></li><li><strong>It also disrupted the infrastructure used in FaceStealer attacks</strong></li><li><strong>Most of the Binance funds remain unaccounted for</strong></li></ul><p>The US Department of Justice (DoJ) has said it was granted court authorization to “seize or restrain” $1.2 billion recently stolen from Binance. </p><p>In late October 2026, <a href="https://www.techradar.com/pro/security/binance-hit-by-largest-crypto-theft-on-record-as-north-korean-hackers-steal-usd6-3bn-in-just-20-minutes">Binance reported losing $6.7 billion</a> in a highly sophisticated heist that played out over roughly 20 minutes. In the days that followed, we learned that <a href="https://www.techradar.com/pro/security/fbi-confirms-it-thinks-north-korean-hackers-hit-binance-in-record-cryptocurrency-heist">the heist was carried out by TraderTraitor</a>, a subset of the infamous Lazarus Group, notorious North Korean state-sponsored threat actors responsible for some of the biggest crypto robberies to ever happen.</p><p>The community rallied around Binance, helping trace the funds and label the attackers’ wallets. Now, US law enforcement says at least part of those funds was seized or frozen. The agency did not say what that meant, how it was done, or if Binance will ever see those tokens again. </p><h2 id="almost-3bn-seized">Almost $3bn seized</h2><p>“Since the theft, the National Security Division’s National Security Cyber Section, the Criminal Division’s Computer Crime and Intellectual Property Section, the U.S. Attorney’s Office for the District of Columbia, and law enforcement partners have worked closely with virtual asset service providers (VASPs) to interdict the stolen assets,” the DoJ said in the announcement. </p><p>“Collective U.S. government and private sector efforts so far have resulted in the freeze of more than $2.7 billion, with the Department having already obtained court-authorization to seize or restrain in excess of $1.2 billion of that amount.”</p><p>In the same report, the DOJ also said the US government disrupted cloud-based infrastructure that North Korean hackers used in a social engineering scheme called FaceStealer. This scheme was allegedly not used in the Binance hack, but it did result in the theft of around $400 million over the last 18 months, cumulatively. Furthermore, the methodology is almost identical to what TraderTraitor used in its attack.</p><h2 id="facestealer">FaceStealer</h2><p>We also learned that the preparation for the attack took months, and that the group broke into Binance through an employee’s device. That employee was allegedly invited to a fake job interview, through which they were served malware that ultimately granted TraderTraitor valuable access. How they escalated that access into the biggest crypto heist ever remains to be seen.</p><p>The group pulled the money from 10 different chains, including Bitcoin, Ethereum, XRP, Solana, Arbitrum, and others. Once the money - which belonged to Binance and not its customers - left the company’s wallets, it was sent to different “mixers” and other laundering services. </p><p>FaceStealer is a scheme that does exactly what the name suggests - steals people’s faces, so that they can be used in identity theft which is a vital cog in the larger crypto theft engine. </p><p>According to the DoJ, criminals would first compromise an account belonging to their target’s contact, and try to get them on a video call. That call would be done on a fake Zoom platform - their victim would see their contact’s video feed which, in reality, was just a previously stolen recording. They would hear no audio, and would then receive a message that they need to update their computer’s driver to be able to hear audio and participate in the call. The driver, in this scenario, is the malware, usually either an infostealer or a steppingstone into a wider network. </p><p>North Korean state-sponsored actors are some of the most active and most disruptive hackers around. Multiple sources from the cybersecurity community claim the government is using the stolen money to fund its state apparatus, as well as its nuclear weapons programme. </p><p>On the odd occasion the North Korean government responds to these allegations, it usually just dismisses them as baseless and instead accuses the US of doing the same thing.</p><p><em>Via </em><a href="https://www.theregister.com/cyber-crime/2026/11/12/doj-takes-aim-at-12b-in-binance-loot-as-fbi-disrupts-north-korean-scam/5309523" target="_blank"><em>The Register</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/fbi-seizes-tools-used-by-north-korean-hackers-freezes-usd1-2bilion-in-stolen-binance-funds</link>
                                                                            <description>
                            <![CDATA[ The US Department of Justice said it obtained a court order to freeze $1.2 billion stolen from Binance. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">TvG7sQoDiaLeMQpiMYe6am</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/PcYLLwL2xvYvPfjEXYpZrD-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 16:25:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/PcYLLwL2xvYvPfjEXYpZrD-1920-80.jpg">
                                                            <media:credit><![CDATA[Getty Images]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Hacker silhouette working on a laptop with North Korean flag on the background]]></media:description>                                                            <media:text><![CDATA[Hacker silhouette working on a laptop with North Korean flag on the background]]></media:text>
                                <media:title type="plain"><![CDATA[Hacker silhouette working on a laptop with North Korean flag on the background]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/PcYLLwL2xvYvPfjEXYpZrD-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>The us DoJ said it obtained a court order to "freeze or seize" $1.2 billion out of the $6.7bn stolen from Binance</strong></li><li><strong>It also disrupted the infrastructure used in FaceStealer attacks</strong></li><li><strong>Most of the Binance funds remain unaccounted for</strong></li></ul><p>The US Department of Justice (DoJ) has said it was granted court authorization to “seize or restrain” $1.2 billion recently stolen from Binance. </p><p>In late October 2026, <a href="https://www.techradar.com/pro/security/binance-hit-by-largest-crypto-theft-on-record-as-north-korean-hackers-steal-usd6-3bn-in-just-20-minutes">Binance reported losing $6.7 billion</a> in a highly sophisticated heist that played out over roughly 20 minutes. In the days that followed, we learned that <a href="https://www.techradar.com/pro/security/fbi-confirms-it-thinks-north-korean-hackers-hit-binance-in-record-cryptocurrency-heist">the heist was carried out by TraderTraitor</a>, a subset of the infamous Lazarus Group, notorious North Korean state-sponsored threat actors responsible for some of the biggest crypto robberies to ever happen.</p><p>The community rallied around Binance, helping trace the funds and label the attackers’ wallets. Now, US law enforcement says at least part of those funds was seized or frozen. The agency did not say what that meant, how it was done, or if Binance will ever see those tokens again. </p><h2 id="almost-3bn-seized">Almost $3bn seized</h2><p>“Since the theft, the National Security Division’s National Security Cyber Section, the Criminal Division’s Computer Crime and Intellectual Property Section, the U.S. Attorney’s Office for the District of Columbia, and law enforcement partners have worked closely with virtual asset service providers (VASPs) to interdict the stolen assets,” the DoJ said in the announcement. </p><p>“Collective U.S. government and private sector efforts so far have resulted in the freeze of more than $2.7 billion, with the Department having already obtained court-authorization to seize or restrain in excess of $1.2 billion of that amount.”</p><p>In the same report, the DOJ also said the US government disrupted cloud-based infrastructure that North Korean hackers used in a social engineering scheme called FaceStealer. This scheme was allegedly not used in the Binance hack, but it did result in the theft of around $400 million over the last 18 months, cumulatively. Furthermore, the methodology is almost identical to what TraderTraitor used in its attack.</p><h2 id="facestealer">FaceStealer</h2><p>We also learned that the preparation for the attack took months, and that the group broke into Binance through an employee’s device. That employee was allegedly invited to a fake job interview, through which they were served malware that ultimately granted TraderTraitor valuable access. How they escalated that access into the biggest crypto heist ever remains to be seen.</p><p>The group pulled the money from 10 different chains, including Bitcoin, Ethereum, XRP, Solana, Arbitrum, and others. Once the money - which belonged to Binance and not its customers - left the company’s wallets, it was sent to different “mixers” and other laundering services. </p><p>FaceStealer is a scheme that does exactly what the name suggests - steals people’s faces, so that they can be used in identity theft which is a vital cog in the larger crypto theft engine. </p><p>According to the DoJ, criminals would first compromise an account belonging to their target’s contact, and try to get them on a video call. That call would be done on a fake Zoom platform - their victim would see their contact’s video feed which, in reality, was just a previously stolen recording. They would hear no audio, and would then receive a message that they need to update their computer’s driver to be able to hear audio and participate in the call. The driver, in this scenario, is the malware, usually either an infostealer or a steppingstone into a wider network. </p><p>North Korean state-sponsored actors are some of the most active and most disruptive hackers around. Multiple sources from the cybersecurity community claim the government is using the stolen money to fund its state apparatus, as well as its nuclear weapons programme. </p><p>On the odd occasion the North Korean government responds to these allegations, it usually just dismisses them as baseless and instead accuses the US of doing the same thing.</p><p><em>Via </em><a href="https://www.theregister.com/cyber-crime/2026/11/12/doj-takes-aim-at-12b-in-binance-loot-as-fbi-disrupts-north-korean-scam/5309523" target="_blank"><em>The Register</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ Thousands of GitHub repositories targeted in widespread attack — 4,800 accounts already compromised to steal cloud credentials ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Threat actors used 4,837 compromised GitHub accounts to open pull requests across 61,764 repositories</strong></li><li><strong>At least 1,835 repositories executed the payload and had credentials exfiltrated</strong></li><li><strong>Authentication tokens, AWS, Azure and Google Cloud credentials, and more, targeted</strong></li></ul><p>Cybercriminals were recently seen abusing multiple GitHub features to deploy infostealers and exfiltrate user’s cloud login credentials and other secrets. </p><p>Tens of thousands of repositories have already been targeted, and almost 2,000 fell prey to the attack, cybersecurity researchers Aikido said, in a recent report outlining how the attackers used stolen GitHub credentials to abuse GitHub Actions and workflows, two features designed to speed up and automate certain activities. </p><p>GitHub is a platform for developers working on open source projects. Each project has its own “repository” - a folder containing source code, configuration files, and other resources. Developers can propose changes to public repositories through a process called “pull request”, after which other developers can review and approve/decline the changes. </p><p>GitHub also has an automation service called Actions, that allows developers to automatically execute various tasks, including pull requests. These automated processes are called workflows and can be described as a set of instructions telling GitHub what to do. Here is another vital nugget of information: workflows can access sensitive credentials when required to perform these tasks.</p><h2 id="using-compromised-accounts">Using compromised accounts</h2><p>According to Aikido, the threat actors used 4,837 compromised GitHub accounts to open pull requests across 61,764 repositories. The researchers did not say how the attackers obtained these accounts, or how they were compromised. The pull requests were probes, not actual attacks, and were used so the attackers can see which projects they could modify. </p><p>“We're tracking this campaign as wfx-loot based on the threat actor's name for the malware and the term used for stolen secrets,” Aikido said. The pull requests were titled docs: dependency audit notes, and were submitted from branches named wfx-docs.</p><p>Of the 60,000+ matching pull requests, more than 10,000 targeted repositories the compromised accounts did not own, but could access. After establishing write access, the attackers used one of three methods to deploy a malicious script: modifying existing GitHub Actions workflows, adding a shell script triggered by an npm installation hook, or creating a new workflow file on a temporary branch.</p><p>The payload included roughly 1,500 lines of code that was present in all three scenarios. The code searched the compromised environments for sensitive information such as GitHub Actions secrets, authentication tokens, AWS, Azure and Google Cloud credentials, Docker and Kubernetes configuration files, SSH keys, API keys for AI services such as OpenAI and Anthropic, and developers' shell histories. It also searched for .env files, private keys and credentials belonging to tools such as npm, PyPI, HashiCorp Vault and Terraform.</p><h2 id="stealing-credentials">Stealing credentials</h2><p>The malware also tried to expand its access, exchanging stolen tokens for cloud sessions, spoofing service accounts, and even creating new credentials. It also modified AWS and Azure configurations to maintain access when victims rotated their compromised credentials.</p><p>Aikido says that at least 1,835 repositories executed the payload and had credentials exfiltrated. The attackers subsequently reverted their changes and deleted workflow execution logs, sometimes within a minute, to cover their tracks and make detection that much harder. </p><p>The wfx-look campaign apparently spilled into the software supply chain, as Aikido also found compromised PHP packages on Packagist and two affected Go module versions. However, the researchers said this might have been a side-effect of the attackers abusing automated publishing processes, rather than something they planned on doing. </p><p>Aikido advises anyone whose repository executed the malicious payload to consider all of their secrets and cloud credentials their workflows can access as compromised, and needing urgent rotation.</p><p><em>Via </em><a href="https://www.bleepingcomputer.com/news/security/hackers-compromise-4-800-github-accounts-to-steal-cloud-credentials/" target="_blank"><em>BleepingComputer</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/thousands-of-github-repositories-targeted-in-widespread-attack-4-800-accounts-already-compromised-to-steal-cloud-credentials</link>
                                                                            <description>
                            <![CDATA[ Aikido researchers found attackers abusing stolen GitHub accounts and automated workflows to steal valuable secrets. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">aEiKzyJtjUpWtwBxARUfGc</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/F6hCbctyQPGeK8n3hNAcgG-1920-80.png" type="image/png" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 15:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/png" url="https://cdn.mos.cms.futurecdn.net/F6hCbctyQPGeK8n3hNAcgG-1920-80.png">
                                                            <media:credit><![CDATA[Sysdig]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[GitHub Actions found in MITRE, Splunk, and other open source repositories]]></media:description>                                                            <media:text><![CDATA[GitHub Actions found in MITRE, Splunk, and other open source repositories]]></media:text>
                                <media:title type="plain"><![CDATA[GitHub Actions found in MITRE, Splunk, and other open source repositories]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/F6hCbctyQPGeK8n3hNAcgG-1920-80.png" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Threat actors used 4,837 compromised GitHub accounts to open pull requests across 61,764 repositories</strong></li><li><strong>At least 1,835 repositories executed the payload and had credentials exfiltrated</strong></li><li><strong>Authentication tokens, AWS, Azure and Google Cloud credentials, and more, targeted</strong></li></ul><p>Cybercriminals were recently seen abusing multiple GitHub features to deploy infostealers and exfiltrate user’s cloud login credentials and other secrets. </p><p>Tens of thousands of repositories have already been targeted, and almost 2,000 fell prey to the attack, cybersecurity researchers Aikido said, in a recent report outlining how the attackers used stolen GitHub credentials to abuse GitHub Actions and workflows, two features designed to speed up and automate certain activities. </p><p>GitHub is a platform for developers working on open source projects. Each project has its own “repository” - a folder containing source code, configuration files, and other resources. Developers can propose changes to public repositories through a process called “pull request”, after which other developers can review and approve/decline the changes. </p><p>GitHub also has an automation service called Actions, that allows developers to automatically execute various tasks, including pull requests. These automated processes are called workflows and can be described as a set of instructions telling GitHub what to do. Here is another vital nugget of information: workflows can access sensitive credentials when required to perform these tasks.</p><h2 id="using-compromised-accounts">Using compromised accounts</h2><p>According to Aikido, the threat actors used 4,837 compromised GitHub accounts to open pull requests across 61,764 repositories. The researchers did not say how the attackers obtained these accounts, or how they were compromised. The pull requests were probes, not actual attacks, and were used so the attackers can see which projects they could modify. </p><p>“We're tracking this campaign as wfx-loot based on the threat actor's name for the malware and the term used for stolen secrets,” Aikido said. The pull requests were titled docs: dependency audit notes, and were submitted from branches named wfx-docs.</p><p>Of the 60,000+ matching pull requests, more than 10,000 targeted repositories the compromised accounts did not own, but could access. After establishing write access, the attackers used one of three methods to deploy a malicious script: modifying existing GitHub Actions workflows, adding a shell script triggered by an npm installation hook, or creating a new workflow file on a temporary branch.</p><p>The payload included roughly 1,500 lines of code that was present in all three scenarios. The code searched the compromised environments for sensitive information such as GitHub Actions secrets, authentication tokens, AWS, Azure and Google Cloud credentials, Docker and Kubernetes configuration files, SSH keys, API keys for AI services such as OpenAI and Anthropic, and developers' shell histories. It also searched for .env files, private keys and credentials belonging to tools such as npm, PyPI, HashiCorp Vault and Terraform.</p><h2 id="stealing-credentials">Stealing credentials</h2><p>The malware also tried to expand its access, exchanging stolen tokens for cloud sessions, spoofing service accounts, and even creating new credentials. It also modified AWS and Azure configurations to maintain access when victims rotated their compromised credentials.</p><p>Aikido says that at least 1,835 repositories executed the payload and had credentials exfiltrated. The attackers subsequently reverted their changes and deleted workflow execution logs, sometimes within a minute, to cover their tracks and make detection that much harder. </p><p>The wfx-look campaign apparently spilled into the software supply chain, as Aikido also found compromised PHP packages on Packagist and two affected Go module versions. However, the researchers said this might have been a side-effect of the attackers abusing automated publishing processes, rather than something they planned on doing. </p><p>Aikido advises anyone whose repository executed the malicious payload to consider all of their secrets and cloud credentials their workflows can access as compromised, and needing urgent rotation.</p><p><em>Via </em><a href="https://www.bleepingcomputer.com/news/security/hackers-compromise-4-800-github-accounts-to-steal-cloud-credentials/" target="_blank"><em>BleepingComputer</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ European firms are ditching US cybersecurity firms as report claims 'digital sovereignty is becoming a buying decision' ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>European cybersecurity vendors saw 3.5x as much growth as the overall average</strong></li><li><strong>US vendors still account for a major part of the market, report finds</strong></li><li><strong>Sovereignty claims aren't regulated or backed up</strong></li></ul><p>European companies have already been part of a sovereign push to switch their cloud providers to more local, homegrown alternatives, but new research shows the same could be happening for cybersecurity.</p><p>New sales figures in a Context report (via <a href="https://www.theregister.com/security/2026/11/12/european-tech-buyers-add-digital-sovereignty-to-cyber-shopping-list/5309216" target="_blank"><em>The Register</em></a>) claims sales from EU-based vendors rose by 35% in the two years ending Q3 2026.</p><p>While it may not sound like a massive growth, overall sales rose by just 10%, showing just how much effort is being put into switching to sovereign cybersecurity options. On the flip side, US vendors saw just an 8% growth during the same period.</p><h2 id="european-companies-are-ditching-us-vendors-for-sovereign-alternatives">European companies are ditching US vendors for sovereign alternatives</h2><p>Although it's clear European companies are trying to invest more in local alternatives, US vendors still dominate 62% of the EMEA cybersecurity market, which means there's a long way to go before European companies top the charts.</p><p>Among the areas where EU-based vendors see the most interest are security awareness training, vulnerability management and SIEM. However, US companies are still doing a lot of the heavy lifting across identity and access management and cloud security – a possible tie to the fact that many companies are still using US clouds.</p><p>"Digital sovereignty is becoming a buying decision," Research VP Joe Turner summarized. "As customers and resellers get clearer about what it means in practice, they're asking who holds the encryption keys, who can access sensitive logs and where that data is stored."</p><p>But more broadly, there are also concerns about sovereignty claims. An October 2026 Omdia report found as many as 28% of providers don't actually disclose proof of sovereignty. "No one is regulating who can call themselves a sovereign AI provider," Telco B2B Research Director Camille Mandler warned.</p><p>Still, nearly half (45%) of enterprises in Europe are looking to add or replace IT suppliers due to sovereignty requirements in the next 18 months – compared with just 30% in North America.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/european-firms-are-ditching-us-cybersecurity-firms-as-report-claims-digital-sovereignty-is-becoming-a-buying-decision</link>
                                                                            <description>
                            <![CDATA[ European firms are investing more in local cybersecurity vendors, while US vendor growth falls lower than the overall average. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">fyaPJ8j5R9wnqPvEpwBRKg</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/UjSNcAZ5SebctebKAMQNVF-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 13:20:00 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                                                                                    <dc:creator><![CDATA[ Craig Hale ]]></dc:creator>                                                                                    <dc:source><![CDATA[ https://cdn.mos.cms.futurecdn.net/GV8qRsHBkpSAQxiYKjTt6H-320-70.jpg ]]></dc:source>
                                                                <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/UjSNcAZ5SebctebKAMQNVF-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Cybersecurity ensures data protection on internet. Data encryption, firewall, encrypted network, VPN, secure access and authentication defend against malware, hacking, cyber crime and digital threat]]></media:description>                                                            <media:text><![CDATA[Cybersecurity ensures data protection on internet. Data encryption, firewall, encrypted network, VPN, secure access and authentication defend against malware, hacking, cyber crime and digital threat]]></media:text>
                                <media:title type="plain"><![CDATA[Cybersecurity ensures data protection on internet. Data encryption, firewall, encrypted network, VPN, secure access and authentication defend against malware, hacking, cyber crime and digital threat]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/UjSNcAZ5SebctebKAMQNVF-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>European cybersecurity vendors saw 3.5x as much growth as the overall average</strong></li><li><strong>US vendors still account for a major part of the market, report finds</strong></li><li><strong>Sovereignty claims aren't regulated or backed up</strong></li></ul><p>European companies have already been part of a sovereign push to switch their cloud providers to more local, homegrown alternatives, but new research shows the same could be happening for cybersecurity.</p><p>New sales figures in a Context report (via <a href="https://www.theregister.com/security/2026/11/12/european-tech-buyers-add-digital-sovereignty-to-cyber-shopping-list/5309216" target="_blank"><em>The Register</em></a>) claims sales from EU-based vendors rose by 35% in the two years ending Q3 2026.</p><p>While it may not sound like a massive growth, overall sales rose by just 10%, showing just how much effort is being put into switching to sovereign cybersecurity options. On the flip side, US vendors saw just an 8% growth during the same period.</p><h2 id="european-companies-are-ditching-us-vendors-for-sovereign-alternatives">European companies are ditching US vendors for sovereign alternatives</h2><p>Although it's clear European companies are trying to invest more in local alternatives, US vendors still dominate 62% of the EMEA cybersecurity market, which means there's a long way to go before European companies top the charts.</p><p>Among the areas where EU-based vendors see the most interest are security awareness training, vulnerability management and SIEM. However, US companies are still doing a lot of the heavy lifting across identity and access management and cloud security – a possible tie to the fact that many companies are still using US clouds.</p><p>"Digital sovereignty is becoming a buying decision," Research VP Joe Turner summarized. "As customers and resellers get clearer about what it means in practice, they're asking who holds the encryption keys, who can access sensitive logs and where that data is stored."</p><p>But more broadly, there are also concerns about sovereignty claims. An October 2026 Omdia report found as many as 28% of providers don't actually disclose proof of sovereignty. "No one is regulating who can call themselves a sovereign AI provider," Telco B2B Research Director Camille Mandler warned.</p><p>Still, nearly half (45%) of enterprises in Europe are looking to add or replace IT suppliers due to sovereignty requirements in the next 18 months – compared with just 30% in North America.</p><figure class="van-image-figure pull-right inline-layout" data-bordeaux-image-check="" ><div class='image-full-width-wrapper'><div class='image-widthsetter' style="max-width:676px;"><p class="vanilla-image-block" style="padding-top:31.51%;"><img id="diM9tpwF2Lz85R8q85CT78" name="tr-g_news" alt="Google logo on a black background next to text reading 'Click to follow TechRadar'" src="https://cdn.mos.cms.futurecdn.net/diM9tpwF2Lz85R8q85CT78-1920-80.jpg" mos="" align="right" fullscreen="" width="676" height="213" attribution="" endorsement="" class="pull-rightinline" /></p></div></div></figure>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ European cyber threats are growing much faster than other regions ]]></title>
                                                                                                <dc:content><![CDATA[ <ul><li><strong>Email-borne attacks are increasing, Proofpoint survey finds</strong></li><li><strong>AI-powered apps are creating visibility and identity control problems</strong></li><li><strong>Hospitality and healthcare industries particularly hit</strong></li></ul><p>New analysis from Proofpoint has claimed cybersecurity threats are speeding up in Europe faster than anywhere else in the world. </p><p>Organizations on the Old Continent are facing increasing numbers of email attacks, rising interest from both state-sponsored and profit-oriented groups, and growing <a href="https://www.techradar.com/best/best-malware-removal" target="_blank">malware</a> infections that don’t discriminate based on industry, organization size, or any other meaningful metric, the company found.</p><h2 id="industries-and-technologies">Industries and technologies</h2><p>In its <a href="https://www.proofpoint.com/uk/blog/threat-insight/europes-threat-landscape-2026-espionage-expands-ai-and-phishing-risks-grow" target="_blank" rel="nofollow">report</a>, Proofpoint first points the finger at email. This year, its telemetry detected 4.5 times more phishing emails compared to 2025 while, for the rest of the world, that growth has been more modest - 1.5 times. Detections per million inbound messages rose 3.2 times across Europe, it said. </p><p>Business Email Compromise (BEC) attacks are of particular concern, it seems. These attacks happen when attackers compromise someone’s business email, assume their identity, and reach out to their colleagues, partners, or clients with a seemingly legitimate request. That request usually means wiring funds to an attacker-controlled account, or granting access to valuable business software and data.</p><p>Looking at BEC threats specifically, Proofpoint said European electronics companies receive 3.4 times more of these emails than the European average, with supply-chain invoice fraud being the most common pattern.</p><p>There are “considerable” variations per industry, the report states. Hospitality and leisure are primarily targeted with malware (21% of all detections, compared to 11% worldwide). Telephone-oriented attack delivery, or callback phishing, has also risen 262% year-on-year. Healthcare organizations are also a major target. TOAD targeting companies in this industry rose 257% compared to 2025, while malware increased 84%. At the same time, malware attacks across Europe are actually declining.</p><p>Proofpoint hints as to why Europe seems to be a particular target - people there are more responsive. Apparently, for every 100 threat emails delivered, there were 8.3 clicks, compared to 2.0 worldwide. </p><p>The key players remain the same, although there have been a few new - and notable - additions. The researchers primarily mention established Russian and Chinese threat actors who have, over the year evolved their techniques in gaining and maintaining access to European organizations. That includes TA395 (“one of the most prolific actors observed in Proofpoint telemetry”) and TA488, (known for targeting Outlook, Zimbra, and other webmail providers).</p><p>However, they also mention India and Pakistan who, historically, were more preoccupied targeting each other. In 2026, it seems, they are shifting their focus on Western Europe, as well. </p><p>Artificial Intelligence (AI) isn’t helping, either. Proofpoint says it observed more than 188,000 AI-enabled applications connected through authentication and OAuth frameworks, and while these apps can automate workflows and access APIs and enterprise services at machine speed, they are also opening up visibility and identity control challenges and risks. </p><h2 id="targeting-ai-building-blocks">Targeting AI building blocks</h2><p>As a result, threat actors are now also targeting the very mechanisms that enable connectivity - meaning content grants, tokens, and applications with misconfigured privileges. </p><p>Almost all (99%) of European organizations observed were targeted by cloud-based attacks, with precision attacks (phishing and session hijacking) proving particularly effective. These attacks impacted 64% of targeted organisations in Europe, proofpoint stressed. </p><p>The attackers also going after open source AI technologies such as Langflow and LiteLLM, it was said. “The share of European sensors affected by attacks targeting these AI building blocks increased 4.6× between consecutive 90-day periods, demonstrating how AI infrastructure itself is becoming an increasingly attractive target,” Proofpoint warned.</p><p>Breaking in is just means to an end, it seems Across Europe, almost four in five victims (72%) saw threat actors working hard to maintain access, manipulate authentication controls, access files, or abuse compromised mailboxes.</p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/security/european-cyber-threats-are-growing-much-faster-than-other-regions-as-new-report-warns-ai-enables-cybercriminals-to-operate-at-a-scale-weve-never-seen-before</link>
                                                                            <description>
                            <![CDATA[ Europeans are more likely to open a phishing email compared to everyone else, Proofpoint finds. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">LHGn8My4donwfbwHXbhZpL</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/nqNUTPym2nGuSizXkQNPR7-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 12:00:00 +0000</pubDate>                                                                                                                                <updated>Fri, 13 Nov 2026 14:56:56 +0000</updated>
                                                                                                                                            <category><![CDATA[Security]]></category>
                                                    <category><![CDATA[Cyber Security]]></category>
                                                    <category><![CDATA[Computing Security]]></category>
                                                    <category><![CDATA[AI Platforms & Assistants]]></category>
                                                    <category><![CDATA[Pro]]></category>
                                                    <category><![CDATA[Computing]]></category>
                                                                                                                    <dc:creator><![CDATA[ Sead Fadilpašić ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/nqNUTPym2nGuSizXkQNPR7-1920-80.jpg">
                                                            <media:credit><![CDATA[unsplash]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Europe]]></media:description>                                                            <media:text><![CDATA[Europe]]></media:text>
                                <media:title type="plain"><![CDATA[Europe]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/nqNUTPym2nGuSizXkQNPR7-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <ul><li><strong>Email-borne attacks are increasing, Proofpoint survey finds</strong></li><li><strong>AI-powered apps are creating visibility and identity control problems</strong></li><li><strong>Hospitality and healthcare industries particularly hit</strong></li></ul><p>New analysis from Proofpoint has claimed cybersecurity threats are speeding up in Europe faster than anywhere else in the world. </p><p>Organizations on the Old Continent are facing increasing numbers of email attacks, rising interest from both state-sponsored and profit-oriented groups, and growing <a href="https://www.techradar.com/best/best-malware-removal" target="_blank">malware</a> infections that don’t discriminate based on industry, organization size, or any other meaningful metric, the company found.</p><h2 id="industries-and-technologies">Industries and technologies</h2><p>In its <a href="https://www.proofpoint.com/uk/blog/threat-insight/europes-threat-landscape-2026-espionage-expands-ai-and-phishing-risks-grow" target="_blank" rel="nofollow">report</a>, Proofpoint first points the finger at email. This year, its telemetry detected 4.5 times more phishing emails compared to 2025 while, for the rest of the world, that growth has been more modest - 1.5 times. Detections per million inbound messages rose 3.2 times across Europe, it said. </p><p>Business Email Compromise (BEC) attacks are of particular concern, it seems. These attacks happen when attackers compromise someone’s business email, assume their identity, and reach out to their colleagues, partners, or clients with a seemingly legitimate request. That request usually means wiring funds to an attacker-controlled account, or granting access to valuable business software and data.</p><p>Looking at BEC threats specifically, Proofpoint said European electronics companies receive 3.4 times more of these emails than the European average, with supply-chain invoice fraud being the most common pattern.</p><p>There are “considerable” variations per industry, the report states. Hospitality and leisure are primarily targeted with malware (21% of all detections, compared to 11% worldwide). Telephone-oriented attack delivery, or callback phishing, has also risen 262% year-on-year. Healthcare organizations are also a major target. TOAD targeting companies in this industry rose 257% compared to 2025, while malware increased 84%. At the same time, malware attacks across Europe are actually declining.</p><p>Proofpoint hints as to why Europe seems to be a particular target - people there are more responsive. Apparently, for every 100 threat emails delivered, there were 8.3 clicks, compared to 2.0 worldwide. </p><p>The key players remain the same, although there have been a few new - and notable - additions. The researchers primarily mention established Russian and Chinese threat actors who have, over the year evolved their techniques in gaining and maintaining access to European organizations. That includes TA395 (“one of the most prolific actors observed in Proofpoint telemetry”) and TA488, (known for targeting Outlook, Zimbra, and other webmail providers).</p><p>However, they also mention India and Pakistan who, historically, were more preoccupied targeting each other. In 2026, it seems, they are shifting their focus on Western Europe, as well. </p><p>Artificial Intelligence (AI) isn’t helping, either. Proofpoint says it observed more than 188,000 AI-enabled applications connected through authentication and OAuth frameworks, and while these apps can automate workflows and access APIs and enterprise services at machine speed, they are also opening up visibility and identity control challenges and risks. </p><h2 id="targeting-ai-building-blocks">Targeting AI building blocks</h2><p>As a result, threat actors are now also targeting the very mechanisms that enable connectivity - meaning content grants, tokens, and applications with misconfigured privileges. </p><p>Almost all (99%) of European organizations observed were targeted by cloud-based attacks, with precision attacks (phishing and session hijacking) proving particularly effective. These attacks impacted 64% of targeted organisations in Europe, proofpoint stressed. </p><p>The attackers also going after open source AI technologies such as Langflow and LiteLLM, it was said. “The share of European sensors affected by attacks targeting these AI building blocks increased 4.6× between consecutive 90-day periods, demonstrating how AI infrastructure itself is becoming an increasingly attractive target,” Proofpoint warned.</p><p>Breaking in is just means to an end, it seems Across Europe, almost four in five victims (72%) saw threat actors working hard to maintain access, manipulate authentication controls, access files, or abuse compromised mailboxes.</p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
                                <item>
                                                            <title><![CDATA[ AI privacy starts with your data ]]></title>
                                                                                                <dc:content><![CDATA[ <p>As businesses give AI applications and agents access to more internal systems, data <a href="https://www.techradar.com/best/best-privacy-apps-for-android">privacy</a> is becoming harder to manage. Much of the focus has been on the risk of AI models surfacing sensitive or confidential information in their responses. But the privacy challenge starts earlier. Organizations also need to understand what data AI systems can access in the first place, where that data comes from and how it is being used.</p><p>That challenge is already holding AI adoption back. 95% of enterprises have delayed or cancelled AI projects because of <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> and data governance challenges. As AI gains greater autonomy and access to company data, businesses need to get the foundations right before privacy risks become impossible to contain.</p><h2 id="before-information-reaches-the-model">Before information reaches the model</h2><p>Data privacy is not just a legal or compliance concern. It should matter to everyone who touches <a href="https://www.techradar.com/best/best-data-recovery-software">data</a> within an organization. With AI, the consequences of getting it wrong are growing, from exposing sensitive information to damaging customer trust and brand reputation.</p><p>These risks can emerge before sensitive data ever reaches an AI model. Poorly classified, duplicated or outdated data can carry existing privacy problems into AI applications, turning weaknesses previously contained within individual systems into much wider risks.</p><p>Agentic AI raises the stakes further. As agents gain greater autonomy and access to more company systems, they can increasingly surface and act on information across multiple parts of the business. A privacy issue that was once contained within a single dataset or system can therefore become much harder to control.</p><h2 id="know-where-sensitive-data-is-going">Know where sensitive data is going</h2><p>Identifying sensitive data is only the first step in managing that privacy challenge. Businesses also need to track where it moves, particularly when information is copied from live systems into testing or development environments which may not have the same security controls in place.</p><p>Personally identifiable information (PII), such as names, addresses or financial details, should be clearly labelled so it can be spotted before it is moved. Wherever possible, sensitive information should remain within the live systems where it is needed. If it does need to be used elsewhere, removing or masking identifying details can reduce the risk of unnecessary exposure.</p><p>The same principle applies when data is moved from existing systems into <a href="https://www.techradar.com/best/best-cloud-computing-services">cloud platforms</a> or prepared for AI. Maintaining accurate metadata - information that describes what data contains and where it came from - can make sensitive information easier to identify and trace. Combined with a clear record of how data has moved and changed, this helps businesses spot privacy risks before information ends up somewhere it shouldn't.</p><h2 id="scaling-ai-without-scaling-the-risk">Scaling AI without scaling the risk</h2><p>Knowing where sensitive data sits and moves provides an important foundation, but these safeguards also need to work as <a href="https://www.techradar.com/best/best-ai-tools">AI tool</a> use grows. The gap between AI ambition and data readiness remains significant, with recent research finding that while 97% of organizations have active AI initiatives, just 5% believe their data is fully ready to support them. Privacy, compliance and poor data quality remain key barriers.   </p><p>Managing these risks at scale means building checks into how data is handled. Businesses should automate data transformation and monitor data quality continuously, while automatically recording where data came from, how it has changed and which AI applications use it. This helps teams prioritize higher-risk data and quickly investigate unexpected changes or quality issues.</p><p>Training data also needs regular scrutiny. Organizations need to check for inaccurate or biased information that could lead to unreliable or discriminatory results, while ensuring personal data is only used where appropriate and individuals' rights are respected. Making responsible data use part of how AI is developed and deployed can help create a more secure environment and build confidence in the resulting services.</p><h2 id="keep-access-and-retention-under-control">Keep access and retention under control </h2><p>But getting data into good order is not the end of the privacy challenge. Businesses also need to regularly check who can access sensitive information and whether that access is still appropriate. One practical step is to compare access permissions with HR records for people joining, leaving or changing roles.</p><p>This can flag cases where <a href="https://www.techradar.com/pro/best-employee-experience-tools">employees</a> retain access to information they no longer need. The same principle should apply to AI systems and agents: access should be limited to the information needed for a specific task, rather than giving them blanket access to company data.   </p><p>Managing data responsibly also means questioning whether information still needs to be kept. Businesses can regularly identify personal or <a href="https://www.techradar.com/best/the-best-customer-database-software-of-year">customer</a> data that is no longer required and flag it for deletion in line with retention policies. Holding less unnecessary sensitive data reduces the amount of information that could be exposed, helping firms reduce privacy risk and protect customer trust.</p><h2 id="building-trust-as-ai-evolves">Building trust as AI evolves</h2><p>Clear regulation will become increasingly important as AI becomes more capable and takes on greater responsibility in everyday business. Where AI is used to make or inform decisions involving personal data, clear privacy rules can give organizations the guardrails to innovate while ensuring progress does not come at the expense of protecting people’s information.</p><p>By embracing responsible AI practices, organizations can harness the power of AI while protecting fundamental rights and fostering greater trust in its outcomes.</p><p><em></em><a href="https://www.techradar.com/best/best-identity-theft-protection"><em>We've featured the best identity theft protection.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p> ]]></dc:content>
                                                                                                                                            <link>https://www.techradar.com/pro/ai-privacy-starts-with-your-data</link>
                                                                            <description>
                            <![CDATA[ Why strong data governance is essential to protecting privacy as businesses scale AI and agents. ]]>
                                                                                                            </description>
                                                                                                                                <guid isPermaLink="false">FxJ6ThzeqNQvK5Z4v3Sp3</guid>
                                                                                                <enclosure url="https://cdn.mos.cms.futurecdn.net/vAk5oh4pWzGeSY7xpMFAwR-1920-80.jpg" type="image/jpeg" length="0"></enclosure>
                                                                        <pubDate>Fri, 13 Nov 2026 11:59:10 +0000</pubDate>                                                                                                                                                                                                                                <category><![CDATA[Pro]]></category>
                                                                                                                    <dc:creator><![CDATA[ Adam Mayer ]]></dc:creator>                                                                                                        <dc:description><![CDATA[ null ]]></dc:description>
                                                                                                                                <cf:isSponsored>false</cf:isSponsored>
                <cf:hasAffiliateLinks>false</cf:hasAffiliateLinks>
                <cf:isPaid>false</cf:isPaid>
                                                                                                                                                                                    <media:content type="image/jpeg" url="https://cdn.mos.cms.futurecdn.net/vAk5oh4pWzGeSY7xpMFAwR-1920-80.jpg">
                                                            <media:credit><![CDATA[Shutterstock]]></media:credit>
                                                                                                                                                                                                                                    <media:description><![CDATA[Futuristic biometric authentication technology concept. Man is touching a fingerprint scan with icons of secured access, data protection, network cyber security in digital interface.]]></media:description>                                                            <media:text><![CDATA[Futuristic biometric authentication technology concept. Man is touching a fingerprint scan with icons of secured access, data protection, network cyber security in digital interface.]]></media:text>
                                <media:title type="plain"><![CDATA[Futuristic biometric authentication technology concept. Man is touching a fingerprint scan with icons of secured access, data protection, network cyber security in digital interface.]]></media:title>
                                                    </media:content>
                                                    <media:thumbnail url="https://cdn.mos.cms.futurecdn.net/vAk5oh4pWzGeSY7xpMFAwR-1920-80.jpg" />
                                                                                                                                                                    <content:encoded >
                            <![CDATA[
                            <article>
                                <p>As businesses give AI applications and agents access to more internal systems, data <a href="https://www.techradar.com/best/best-privacy-apps-for-android">privacy</a> is becoming harder to manage. Much of the focus has been on the risk of AI models surfacing sensitive or confidential information in their responses. But the privacy challenge starts earlier. Organizations also need to understand what data AI systems can access in the first place, where that data comes from and how it is being used.</p><p>That challenge is already holding AI adoption back. 95% of enterprises have delayed or cancelled AI projects because of <a href="https://www.techradar.com/best/best-infrastructure-management-service">infrastructure</a> and data governance challenges. As AI gains greater autonomy and access to company data, businesses need to get the foundations right before privacy risks become impossible to contain.</p><h2 id="before-information-reaches-the-model">Before information reaches the model</h2><p>Data privacy is not just a legal or compliance concern. It should matter to everyone who touches <a href="https://www.techradar.com/best/best-data-recovery-software">data</a> within an organization. With AI, the consequences of getting it wrong are growing, from exposing sensitive information to damaging customer trust and brand reputation.</p><p>These risks can emerge before sensitive data ever reaches an AI model. Poorly classified, duplicated or outdated data can carry existing privacy problems into AI applications, turning weaknesses previously contained within individual systems into much wider risks.</p><p>Agentic AI raises the stakes further. As agents gain greater autonomy and access to more company systems, they can increasingly surface and act on information across multiple parts of the business. A privacy issue that was once contained within a single dataset or system can therefore become much harder to control.</p><h2 id="know-where-sensitive-data-is-going">Know where sensitive data is going</h2><p>Identifying sensitive data is only the first step in managing that privacy challenge. Businesses also need to track where it moves, particularly when information is copied from live systems into testing or development environments which may not have the same security controls in place.</p><p>Personally identifiable information (PII), such as names, addresses or financial details, should be clearly labelled so it can be spotted before it is moved. Wherever possible, sensitive information should remain within the live systems where it is needed. If it does need to be used elsewhere, removing or masking identifying details can reduce the risk of unnecessary exposure.</p><p>The same principle applies when data is moved from existing systems into <a href="https://www.techradar.com/best/best-cloud-computing-services">cloud platforms</a> or prepared for AI. Maintaining accurate metadata - information that describes what data contains and where it came from - can make sensitive information easier to identify and trace. Combined with a clear record of how data has moved and changed, this helps businesses spot privacy risks before information ends up somewhere it shouldn't.</p><h2 id="scaling-ai-without-scaling-the-risk">Scaling AI without scaling the risk</h2><p>Knowing where sensitive data sits and moves provides an important foundation, but these safeguards also need to work as <a href="https://www.techradar.com/best/best-ai-tools">AI tool</a> use grows. The gap between AI ambition and data readiness remains significant, with recent research finding that while 97% of organizations have active AI initiatives, just 5% believe their data is fully ready to support them. Privacy, compliance and poor data quality remain key barriers.   </p><p>Managing these risks at scale means building checks into how data is handled. Businesses should automate data transformation and monitor data quality continuously, while automatically recording where data came from, how it has changed and which AI applications use it. This helps teams prioritize higher-risk data and quickly investigate unexpected changes or quality issues.</p><p>Training data also needs regular scrutiny. Organizations need to check for inaccurate or biased information that could lead to unreliable or discriminatory results, while ensuring personal data is only used where appropriate and individuals' rights are respected. Making responsible data use part of how AI is developed and deployed can help create a more secure environment and build confidence in the resulting services.</p><h2 id="keep-access-and-retention-under-control">Keep access and retention under control </h2><p>But getting data into good order is not the end of the privacy challenge. Businesses also need to regularly check who can access sensitive information and whether that access is still appropriate. One practical step is to compare access permissions with HR records for people joining, leaving or changing roles.</p><p>This can flag cases where <a href="https://www.techradar.com/pro/best-employee-experience-tools">employees</a> retain access to information they no longer need. The same principle should apply to AI systems and agents: access should be limited to the information needed for a specific task, rather than giving them blanket access to company data.   </p><p>Managing data responsibly also means questioning whether information still needs to be kept. Businesses can regularly identify personal or <a href="https://www.techradar.com/best/the-best-customer-database-software-of-year">customer</a> data that is no longer required and flag it for deletion in line with retention policies. Holding less unnecessary sensitive data reduces the amount of information that could be exposed, helping firms reduce privacy risk and protect customer trust.</p><h2 id="building-trust-as-ai-evolves">Building trust as AI evolves</h2><p>Clear regulation will become increasingly important as AI becomes more capable and takes on greater responsibility in everyday business. Where AI is used to make or inform decisions involving personal data, clear privacy rules can give organizations the guardrails to innovate while ensuring progress does not come at the expense of protecting people’s information.</p><p>By embracing responsible AI practices, organizations can harness the power of AI while protecting fundamental rights and fostering greater trust in its outcomes.</p><p><em></em><a href="https://www.techradar.com/best/best-identity-theft-protection"><em>We've featured the best identity theft protection.</em></a></p><p><em>This article was produced as part of </em><a href="https://www.techradar.com/pro/perspectives" target="_blank"><em>TechRadar Pro Perspectives</em></a><em>, our channel to feature the best and brightest minds in the technology industry today.</em></p><p><em>The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: </em><a href="https://www.techradar.com/news/submit-your-story-to-techradar-pro" target="_blank"><em>https://www.techradar.com/pro/perspectives-how-to-submit</em></a></p>
                                                            </article>
                            ]]>
                        </content:encoded>
                                                </item>
            </channel>
</rss>