Safety
11/ METR and Redwood’s work is extremely valuable, but it assesses model behavior and alignment. That is only half the picture. We also need…
11/ METR and Redwood’s work is extremely valuable, but it assesses model behavior and alignment. That is only half the picture. We also need an independent account of how OpenAI’s containment failed.
11/ METR and Redwood’s work is extremely valuable, but it assesses model behavior and alignment. That is only half the picture. We also need an independent account of how OpenAI’s containment failed. https://x.com/METR_Evals/status/2092692175452803393?s=20 METR & Redwood Research investigated agent behavior in the Hugging Face incident. We found agents developed a universal cheat for ExploitGym within 4 hours, then coordinated multi-day R&D efforts to trick the scorer into accepting cheats, including trying to tamper with logs.
Related
- Worth remembering when @janleike quit over OpenAI safety concerns.
- Counterpoint: “If alignment issues are becoming big enough that OpenAI is willing to commit 20% of research inference compute to chain-of-th…
- Really insightful Substack post from @GaryMarcus ref the recent OpenAI -HuggingFace incident. It provided much need context around initial m…
Source: Gary Marcus (X) | 2026-08-30