Become a TechRadar Insider
- OpenAI has unveiled Daybreak, its latest security project
- It seeks to rival Anthropic's Mythos in detecting and patching high-severity vulnerabilties
- Daybreak will also help companies build software more securely from the start
OpenAI has unveiled Daybreak, its answer to Anthropic’s Mythos and Project Glasswing, sparking a potential cybersecurity arms race between the two companies.
“Daybreak combines the intelligence of OpenAI models, the extensibility of Codex as an agentic harness, and our partners across the security flywheel to help make the world safer for everyone,” the announcement said.
The project looks to work with OpenAI’s industry and government partners by securing software at the very beginning of the development process, creating a more robust base that, in time, will scale the effectiveness of cyber defense.
Daybreak to build software securely from the start
In the blog post, OpenAI sums up the project in a single sentence: “The goal is simple: accelerate cyber defenders and continuously secure software.”
Daybreak will allow organizations to apply OpenAI’s Codex Security to their own repository using an ‘agentic harness’, where it will seek out, analyze, and patch attack paths and high-impact code.
High-priority vulnerabilities can be analyzed and validated in a secure, isolated environment, “so teams can prioritize real, reproducible issues over noisy alerts.” Codex Security will also allow teams to automate detection and response, increasing efficiency and securing critical vulnerabilities faster.
Daybreak therefore seeks to delegate the rote work of identifying and analyzing to AI, and to return the evidence-backed results of vulnerabilities to human teams. Where Daybreak differs in its approach compared to Mythos is in building software securely from the start and constantly monitoring for vulnerabilities, compared to Mythos’ focus of detecting and mitigating high-severity vulnerabilities at scale.
Daybreak includes three models; GPT-5.5, as the default for general-purpose work; GPT-5.5 with Trusted Access for Cyber, to be used in defensive security workflows; and GPT-5.5-Cyber, for specialized workflows including red teaming and pen testing.
Dane Knecht, CTO at Cloudflare, said, “We’re excited about the potential of OpenAI’s cyber capabilities to bring stronger reasoning and more agentic execution into security workflows. It’s a big step forward for teams to be able to leverage frontier models not only to accelerate velocity, but also to improve their security posture.”
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews, and opinion in your feeds.
