OpenAI Launches Daybreak Red Team and Defense Factory with GPT-6 Cyber Agents
Summary
OpenAI releases Daybreak Red Team, an early-access penetration-testing tool, and Daybreak Defense Factory, which uses GPT-6 Cyber agents to find and patch code vulnerabilities. President Greg Brockman claims companies are secure only to the extent of the frontier model they use. Nonprofits sue in San Francisco Superior Court to halt more capable AI systems.
Key Points
- Nancy Wang says Daybreak Red Team offers 'superhuman pen-test at the scale of automated continuous scanning' with fleets of agents.
- Engineers approve Daybreak Defense Factory patches before production after review, says OpenAI's Nancy Wang.
- Daybreak Red Team tests patches for flaws reported by code scanners and bug bounty programs.