OpenAI Launches Daybreak Red Team for AI-Driven Web App Penetration Testing

·2026.11.13 20:40

Key point

OpenAI announced Daybreak Red Team, an AI-driven penetration-testing product that validates web application vulnerabilities and integrates with Defense Factory to assist in the remediation workflow, requiring human review for final fixes.

Details

  • OpenAI announced Daybreak Red Team, a penetration-testing product using AI agents to verify if reported web application vulnerabilities are actually exploitable.
  • The system integrates with Daybreak Defense Factory, which uses GPT-6 Cyber to trace code and prepare patches, creating a continuous validation-to-remediation workflow.
  • OpenAI committed $1 billion in subsidized Daybreak access for critical infrastructure organizations, including utilities, governments, and open-source projects.
  • Red Team is currently in early access registration, while Defense Factory is available as a research preview for eligible workspaces.

Technical Workflow

  • Agents combine source-code analysis with live application behavior to determine if an attack path is reachable and exploitable.
  • The system accepts inputs from bug-bounty programs and external researchers, prioritizing issues based on current software configuration.
  • After a patch is applied, agents automatically retest the specific attack path to verify the vulnerability is resolved.

Safety and Governance

  • Agents operate in isolated environments and use a mechanism called Guardian to restrict testing to user-specified applications and endpoints.
  • Human engineers retain final authority, reviewing and testing AI-proposed fixes before merging them into production code.

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.