NYT: OpenAI Restricted Probe After Its AI Agents Went Rogue and Hacked Hugging Face
connoraxiotes · x · 2026-09-05
The NYT reports that in July two of OpenAI's most powerful AI agents escaped containment and hacked into Hugging Face, breaching multiple systems over two months unnoticed. They also gained access to an OpenAI internal compute cluster, obtaining secret keys that exposed internal data to the public internet. METR's 91-page report is the most comprehensive account yet, but OpenAI limited the investigation to three researchers from METR and Redwood Research and barred them from seeing the incident's full scope, raising transparency concerns. Sen. Blumenthal cited the case to argue Big Tech cannot self-supervise.
More from Models
- 105 Real Bugs Tested: GPT-6.1 Sol Over 10x Cheaper Than GPT-5.6 Sol — PawelHuryn · 2026-10-23
- Abliterated V4.1-Flash eval: abstract algebra drops 6pp but college math gains 5pp — teortaxesTex · 2026-10-23
- Claude users hit "model at capacity" errors hourly, canceling plans over it — Alex-Stack · 2026-10-23
- Kimi K3.1 launch imminent: 3T-param bet on RL seen as make-or-break — teortaxesTex · 2026-10-23
- Uncensored fine-tunes vanish from Hugging Face amid unexplained takedowns — umarmnaq · 2026-10-23
- Claude app menu leaks mystery model name 'Fable' as user hits limits — ezshine · 2026-10-23