AI tools · November 13, 2026

OpenAI's Daybreak agents will pen-test your app — and then write the patch

OpenAI is now selling AI agents that attack your software and then fix what they find. At its November 12 “Cyber at the Frontier” event, the company launched two products under its Daybreak security brand: Daybreak Red Team, which sends coordinated agent teams to penetration-test web applications, and Daybreak Defense Factory, which turns the findings into tested patches.

Red Team works like a tireless junior pen-tester: its agents use source code and application context to explore attack paths, check whether reported flaws can actually be exploited, validate findings from scanners, bug bounties, and disclosure reports, and retest the app after a fix lands. Defense Factory picks up the other half of the job — it takes findings from tools like Wiz, Snyk, and Endor Labs, plus connected GitHub repositories, traces each vulnerability through the code, and prepares a patch as a draft pull request. The agents run on OpenAI's general models and its Daybreak Blue and Daybreak Red security models.

The important guardrail is human: engineers review and test every proposed fix before it merges, and the platform does not deploy patches on its own. Red Team is in early access by request, and Defense Factory is sold through OpenAI's sales team; neither product has public pricing. OpenAI has published no independent results for detection rates, false positives, or patch quality, so teams considering it should pilot on a non-production repo first and measure their own mean time from finding to merged fix.

Why it matters

Security teams drown in scanner alerts they never have time to verify, and most known vulnerabilities sit unpatched for months. Agents that prove a flaw is real and hand over a tested fix — with a human still holding the merge button — attack the backlog itself, not just the alert count.

Key facts

Announced November 12, 2026, at OpenAI's “Cyber at the Frontier” event, under the Daybreak security brand
Daybreak Red Team is in early access: coordinated agents explore web-app attack paths using source code and application context, validate scanner and bug-bounty findings, and retest apps after patches
Daybreak Defense Factory takes findings from Wiz, Snyk, and Endor Labs plus connected GitHub repos, and has agents trace each issue through the code and prepare a patch
Patches arrive as draft pull requests with tests; engineers review and test every change before merging — nothing is deployed automatically
The agents run on OpenAI's general models plus its Daybreak Blue and Daybreak Red security models, according to OpenAI's product pages
No public pricing for either product; Red Team is by early-access request and Defense Factory through sales

Sources

Spot an error? We correct quickly and note it. Contact us via the contact page.

← More AI news