OpenAI launches AI tools to test and fix software flaws

OpenAI announced two Daybreak cybersecurity products at its Cyber at the Frontier event on 12 November. Daybreak Red Team uses AI agents to test web applications, investigate potential weaknesses and validate vulnerabilities. OpenAI says the agents use source code and application context to trace attack paths, and can assess findings from code scanners, bug bounty submissions and vulnerability disclosure reports. They work in isolated environments, with a Guardian review process restricting which applications and endpoints they can test.

Daybreak Defense Factory is designed to investigate security findings and prepare proposed patches. It can take findings from Red Team and other products, including imported vulnerability data from Wiz, Snyk and Endor Labs. The system can scan connected GitHub repositories, monitor commits and run scheduled assessments.

Engineers can review evidence, modify proposed fixes using Codex and create draft GitHub pull requests; patches are not deployed automatically, and engineers must review and test them before merging.

OpenAI is accepting requests for early access to Daybreak Red Team. Defense Factory is available as a preview plugin in ChatGPT workspaces with Daybreak access.