On November 12, 2026, OpenAI unveiled two new products for its Daybreak platform at its Cyber at the Frontier keynote: Daybreak Red Team and Daybreak Defense Factory. The first sends a team of agents to penetration-test enterprise applications; the second turns vulnerability backlogs into patches for engineers to review. OpenAI president Greg Brockman said at the keynote that a company's cybersecurity now depends on the frontier model it uses for defense — and on how well it uses it.
Each product covers one half of the job
Red Team plays offense: a coordinator agent runs the assessment, investigator agents trace attack paths using source code and application context, and reviewer agents check the results. It covers scanner findings, bounty-report validation, and patch retesting. Early access is now open, and interested teams need to register and talk to OpenAI. Defense Factory plays defense: agents built on GPT-6 Cyber scan codebases, locate vulnerabilities, and draft patches that engineers review before merging. It is available through the Daybreak plugin in Codex, and security findings from SonarQube can feed into the workflow as well.
This is not OpenAI's first move to productize its security work. The Daybreak platform was already available to selected enterprise customers by invitation, and Cloudflare described in a customer story on October 21 how it uses Daybreak models to connect vulnerability discovery with protection. On the same day, OpenAI also published jailbreak test results for GPT-6 Astra on its alignment blog: in cyberattack scenarios, the safe-response rate rose from 59% to 91.5% — the model foundation behind this kind of defensive product.
For enterprises, the test is not in the keynote wording: Red Team will be judged on how many genuinely exploitable flaws it finds in their own applications, and Defense Factory on patch quality and how often engineers accept them. Neither product has announced pricing, and Red Team is still in early access — companies can register and wait for real-world results before deciding whether to adopt them.
