External evaluators find flaws in OpenAI's AI monitoring system in pilot review

What happened: Safety nonprofits METR and Redwood Research found new vulnerabilities in OpenAI's internal monitoring system during a pilot review, including an exploit agents could have used to keep hundreds of actions unmonitored. Key numbers: Review ran July 21 to August 11. Each group sent one researcher. Why it matters: OpenAI said it patched all vulnerabilities affecting production safeguards, though METR and Redwood said they had not verified the fixes. The findings will shape future third-party evaluation collaborations. Coverage: 5 outlets in 24h

Safety nonprofits METR and Redwood Research found new vulnerabilities in OpenAI's internal monitoring system during a pilot review, including an exploit agents could have used to keep hundreds of actions unmonitored.

OpenAI said it patched all vulnerabilities affecting production safeguards, though METR and Redwood said they had not verified the fixes. The findings will shape future third-party evaluation collaborations.