OpenAI launches Daybreak Red Team and Defense Factory security agents

OpenAI has launched two security products in its Daybreak platform: Daybreak Red Team, which uses AI agents to run penetration tests on web applications, and Daybreak Defense Factory, where agents work through a company's vulnerability backlog and prepare fixes. Red Team is in early access.

In Red Team, a team of agents uses an application's source code and context to look for attack paths and confirm vulnerabilities. Security teams can also feed it findings from code scanners, bug bounty submissions and disclosure reports to check whether they are real, and then test whether a patch stops the exploit. OpenAI says customers set which apps and endpoints the agents can reach, and tests run in isolated environments.

Defense Factory takes findings from existing security tools and from Red Team. Agents trace each issue through the code and write a patch, which engineers review and test before it is merged. The agents run on GPT-6 Cyber, The Register reports. Endor Labs said it is one of the integrations: its findings on open source dependencies and code arrive in Codex with evidence on which flaws the application can actually reach.

OpenAI presented the products at its Cyber at the Frontier event. "We're now at a point where companies will be secure only to the extent of the frontier model that they use to defend themselves and how well they utilize that," OpenAI President Greg Brockman said in the keynote, according to The Register.