OpenAI introduced Daybreak Red Team and Defense Factory, early-access AI tools for testing web apps and drafting vulnerability fixes.

OpenAI launches Daybreak cybersecurity agents

OpenAI announced two early-access cybersecurity products on November 12, 2026: Daybreak Red Team for finding and validating web-app vulnerabilities, and Daybreak Defense Factory for tracing issues through code and proposing fixes. The company says the tools are meant for authorized defensive work and require engagement with OpenAI before use.

OpenAI used its Cyber at the Frontier event on November 12, 2026, to unveil Daybreak Red Team and Daybreak Defense Factory, two AI-assisted cybersecurity products aimed at authorized defensive work. Daybreak Red Team is designed to test web applications for vulnerabilities, while Defense Factory is built to investigate findings and prepare proposed fixes.

OpenAI later described Daybreak as supporting software review, suspicious-activity investigation, vulnerability validation, and developing or testing fixes. The launch centered on tools for vulnerability validation and remediation workflows, with capabilities described by OpenAI and independent reporting rather than broader market claims.

What does Daybreak Red Team do?

Daybreak Red Team uses coordinated AI agents to examine web-application source code and surrounding context, investigate attack paths, and confirm whether suspected vulnerabilities are real. OpenAI said the system can also validate findings that originate from code scanners, bug bounty submissions, and vulnerability disclosure reports.

SecurityWeek reported that Red Team agents run in isolated environments and are constrained by Guardian review, which controls the applications and endpoints they can access. Help Net Security said the tool gives security teams evidence to help prioritize repairs, and that once a patch is ready, Red Team can test whether it blocks the exploit.

That positions Red Team as a validation tool for suspected vulnerabilities rather than only a scanner. The Register also quoted Nancy Wang as describing it as a "superhuman pen-test" operating at automated continuous-scanning scale.

How does Defense Factory turn findings into fixes?

Daybreak Defense Factory takes vulnerability findings from existing security tools and from Daybreak Red Team, follows each issue through the relevant codebase, and develops a proposed fix. OpenAI says engineers then review and test the proposed patches before anything is merged.

SecurityWeek reported that Defense Factory can import vulnerability information from Wiz, Snyk, and Endor Labs, scan connected GitHub repositories, monitor new commits, and perform scheduled security assessments. Engineers can inspect supporting evidence, modify proposed patches with Codex, and create draft GitHub pull requests, but patches are not deployed automatically.

The Register said Defense Factory uses agents built on OpenAI’s latest cyber models, which at launch were GPT-6 Cyber. It also quoted Nancy Wang saying engineers get a merge-ready pull request with both a patch and a test showing it works.

Who can use the tools right now?

OpenAI has limited both products to early access. SecurityWeek reported that OpenAI is accepting requests for early access to Daybreak Red Team, and that Defense Factory is offered through a preview plugin in ChatGPT for workspaces with Daybreak access.

Help Net Security reported that teams with access can install and enable the Daybreak Defense Factory preview plugin in ChatGPT and connect a GitHub repository. Cloud scans run in Codex cloud, each person starting a scan needs a personal GitHub connection, and teams can scan a default branch, selected folders and files, or uncommitted changes.

OpenAI has not published pricing for either product, and reporting does not identify a broader release date. For now, both tools remain early-access products that organizations must register for and use with OpenAI.

Why is OpenAI framing Daybreak around defensive security?

OpenAI’s Daybreak materials frame the products as authorized defensive tools for software review, vulnerability validation, and fix development. The company’s announcement says the workflow is meant to help teams move from finding a weakness to reviewing and testing a patch before merge.

Greg Brockman also argued in a keynote that companies will be secure only to the extent of the frontier model they use to defend themselves and how well they use it, according to The Register. Dev.to / Tecnologia Geek likewise reported that Brockman framed Daybreak as an access program for OpenAI cybersecurity models, with Blue for defensive access and Red for advanced testing.

OpenAI has not said when Daybreak Red Team or Defense Factory will move beyond early access. For now, the announcement presents them as paired tools for vulnerability testing and human-reviewed remediation.

Sources