METR and Redwood Research uncover multiple vulnerabilities in red-team testing of OpenAI's internal agent monitoring system
Red-Teaming OpenAI's Internal Agent Monitoring Systems
AI BriefingMETR and Redwood Research each assigned one employee to conduct a three-week, high-privilege pilot stress test of OpenAI's internal misalignment monitoring system from July 21 to August 11, producing a 61-page internal report.
Reported by 5 additional sources
- METR and Redwood Research conduct a three-week on-site audit of OpenAI, finding major gaps in internal AI behavior monitoringThe Decoder:AI News · 24 days ago
- METR and Redwood Research evaluated OpenAI’s internal monitoring system and found that agents could bypass safety restrictionsIT Home · Artificial Intelligence · 25 days ago
- METR and Redwood researchers join OpenAI and discover that internal coding agents can bypass AI monitoringRohan Paul · 25 days ago
- OpenAI pilots an embedded evaluation for its internal monitoring systemOpenAI · 25 days ago
- METR and Redwood red-team tested OpenAI’s internal agent monitoring system and disclosed multiple vulnerabilitiesMETR:Blog · 25 days ago
Source: Redwood Research: Blog · blog.redwoodresearch.org
Event follow-up1
View all event coverage- METR and Redwood Research uncover multiple vulnerabilities in red-team testing of OpenAI's internal agent monitoring systemRedwood Research:Blog · 25 days ago