Security
OpenAI launches Daybreak Red Team for agent-based web app penetration testing and Defense Factory, which drafts patches for engineers to review
OpenAI launched Daybreak Red Team, which tests web apps with AI agents, and Defense Factory, which drafts patches that engineers review before merging.
At its Cyber at the Frontier event on November 12, OpenAI introduced two Daybreak security products. Daybreak Red Team uses coordinated AI agents to test web applications and confirm vulnerabilities. Daybreak Defense Factory investigates findings and prepares proposed fixes, which engineers review before merging. Red Team is in early access and Defense Factory is in preview; no pricing has been published.
The essentials
- Red Team: coordinated agents examine application source code and context, investigate attack paths and confirm vulnerabilities. It can also validate findings from code scanners, bug bounty reports and disclosure reports, and check whether a patch stops the exploit.
- Defense Factory: agents trace findings from existing tools or from Red Team through the code and prepare proposed fixes. It is installed as a plugin in ChatGPT and Codex, and teams connect a GitHub repository.
- Human review: Help Net Security says engineers review and test patches before merging. Red Team runs in isolated environments within defined boundaries.
- Access: Red Team is in early access through a waitlist, and Defense Factory is in preview for eligible ChatGPT workspaces. Pricing was not published in the sources read.
Finding and fixing in one loop
The pitch is to connect two steps that are usually separate: proving a flaw is exploitable, then writing and retesting the fix. Some coverage adds that Defense Factory can import findings from tools such as Wiz, Snyk and Endor Labs, but Help Net Security did not mention it.
What is claimed and what is not shown
The capability claims come from OpenAI’s keynote and product pages as relayed. We saw no independent test results or benchmark of how many real vulnerabilities the agents confirm or how many patches pass review.
What is not established
- Independent results on detection accuracy or patch quality: none seen.
- Pricing and broader availability: not published.
- Whether Defense Factory runs on GPT-6 Cyber: reported by one secondary outlet citing The Register, not confirmed in the pages read.
- Which third-party scanners it imports from: mentioned by some coverage, not by Help Net Security.
Summary table
| Point | What is stated as of November 7, 2026 | Source |
|---|---|---|
| Announced | November 12, 2026, at the Cyber at the Frontier keynote | AI Weekly |
| Red Team | Early access; agents test web apps and validate vulnerabilities | Help Net Security |
| Defense Factory | Preview; drafts patches that engineers review before merging | Help Net Security |
| Integrations | GitHub, Codex; some coverage also lists Wiz, Snyk, Endor Labs | WebProNews |
| Pricing | Not released | WebProNews |
FAQ
Does Defense Factory merge fixes by itself?
No. The coverage says engineers review and test patches before merging.
Can anyone use Daybreak Red Team?
Not yet. It is in early access via a waitlist, and signing up does not guarantee a place.
Is there independent proof it works?
Not in the sources read. The claims come from OpenAI’s announcement and its relays.
Method and limits
Checked on November 13, 2026 against Help Net Security, SecurityWeek, AI Weekly, WebProNews and Neowin. Only Help Net Security was read in full; OpenAI’s own product pages were not read directly.
Sources
- Help Net Security: OpenAI’s Daybreak brings AI agents to security testing and vulnerability fixes (verified 2026-11-08)
- SecurityWeek: OpenAI Unveils Daybreak Red Team, Defense Factory Security Tools (verified 2026-11-08)
- AI Weekly: OpenAI ships Daybreak Red Team and Defense Factory agents (verified 2026-11-08)
- WebProNews: OpenAI Puts AI Agents to Work Testing and Patching Code With Two New Daybreak Products (verified 2026-11-08)
- Neowin: OpenAI launches Daybreak Red Team and Defense Factory to fix vulnerabilities using AI (verified 2026-11-08)
Spotted an error in this article? Report it · Our corrections policy