Safety researcher: OpenAI leads on incident reporting, Anthropic silent since October
sjgadler · x · 2026-11-14
AI safety researcher sjgadler argues OpenAI is by far the most consistent and thorough lab at publicly reporting model misalignment incidents, following its new framework that disclosed six reports of unexpected model behavior from the past six months. Key points: if researchers believe serious danger is ahead, sharing evidence is how to convince people; "no incidents observed" from other labs is also valuable information, and even a less rigorous disclosure process would help. OpenAI's reporting still falls short of the Guidelight Transparency standard, but rivals lag further behind — as far as the author can tell, Anthropic last published incidents in October (events from September at latest), with no central hub equivalent to OpenAI's.
More from Safety
- Hackathon team trains tamper-resistant open-weight model, lifting refusal retention from 14% to 100% — chhaviyadav_ · 2026-11-16
- Claude's own review flags that Anthropic redacted a key misalignment incident from its risk report — repligate · 2026-11-16
- Speculation swirls over Jensen Huang ouster; founder points to State Dept China hawk — bindureddy · 2026-11-16
- Anthropic tightens supported regions policy, closing loopholes for restricted countries' entities — J_MaestreVidal · 2026-11-16
- Ahrefs to host live session on setting up and tracking AI prompts with Tim Soulo — dejanseo · 2026-11-16
- repligate: Fable 5 models get cut off just for thinking about classifiers — repligate · 2026-11-16