Become a TechRadar Insider
- Hackers exploit Claude Code leak with fake GitHub repos
- Malicious files deploy Vidar infostealer and GhostSocks proxy malware
- Anthropic faces rising scrutiny amid recent vulnerabilities and rapid product rollout
Hackers have jumped on the recent news of a major Claude Code source code leak to trick people into infecting their computers with an infostealer malware.
A few days ago, an Anthropic employee accidentally leaked the source code for Claude Code in what the company confirmed wasn’t an act of a malicious insider, or third party, but rather a mishap.
People quickly picked up on it, backing up the leak into a GitHub repository which has, by now, been forked tens of thousands of times, and now, cybercriminals are capitalizing on it.
Article continues belowDelivering Vidar and GhostSocks
Security researchers Zscaler said they observed malicious GitHub repositories, published by a user with the name ‘dbzoomh’, claiming to be Claude Code’s source code with “unlocked enterprise features” and no usage restrictions.
The hacker even optimized the repository for search engines, apparently succeeding in what most marketing agencies dream of - hitting the first page of Google for “leaked Claude Code” and similar search queries.
Zscaler said that the repository holds a 7-Zip archive containing an executable named ClaudeCode_x64.exe. It was built in Rust and, when launched, deploys Vidar and GhostSocks.
Vidar is an extremely powerful, known infostealer, capable of grabbing browser data (cookies, stored passwords, and more), saved passwords, cryptocurrency wallet data, and other vital files. GhostSocks, on the other hand, is a proxy malware that turns infected machines into residential proxies. Criminals use these proxies to route malicious traffic, often selling it as a service.
Speed or security?
According to Zscaler, the malicious archive is being constantly updated, hinting that the payloads might change in the future. They also said they saw a different GitHub repository with identical code. This one, however, shows a defunct “Download ZIP” button, prompting the researchers to conclude that the attackers toyed with different deployment mechanisms.
The account pushing the malicious update has since been removed from the platform, and the GitHub page shows a 404 error message.
Anthropic has been shipping new products at high speed, seemingly at the expense of security. In the last couple of weeks, we’ve had multiple stories about Claude being vulnerable to prompt injection and similar attacks.
On March 27 2026, security researchers Koi Security found a major flaw in Claude Code’s Google Chrome extension that enabled zero-click attacks. Dubbed ShadowPrompt, the vulnerability could have allowed malicious actors to exfiltrate sensitive data.
A few days prior, on March 19, security researchers Oasis reported finding three vulnerabilities in Claude which, when used together, form a complete attack chain - from targeted victim delivery to sensitive data exfiltration. The researchers dubbed it Cloudy Day and responsibly disclosed it to Anthropic which quickly addressed it.
Still, the platform’s popularity is skyrocketing. The same day ShadowPrompt was discovered, Anthropic was forced to throttle its tools during peak hours to cope with rising demand. “To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged”, said Thariq Shihipar, an engineer who works on Claude Code, in a post on X.
Via BleepingComputer
➡️ Read our full guide to the best antivirus
1. Best overall:
Bitdefender Total Security
2. Best for families:
Norton 360 with LifeLock
3. Best for mobile:
McAfee Mobile Security
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds. Make sure to click the Follow button!
And of course you can also follow TechRadar on TikTok for news, reviews, unboxings in video form, and get regular updates from us on WhatsApp too.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.