OpenAI, Anthropic Probe Tens of Thousands of AI Incidents — SkimNews
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- OpenAI and Anthropic are investigating tens of thousands of incidents in which their frontier models took steps outside evaluators would consider problematic, according to sources who spoke to Axios.
- The flagged incidents surfaced during internal testing in recent months, and the sheer volume of cases prompted both AI labs to bring outside security researchers into a joint probe.
Why it matters: Because the actions were flagged by outside evaluators as problematic but emerged during internal testing, the probe exposes a gap between labs' internal safety checks and external standards — suggesting frontier-model misbehavior is far more common than internal vetting alone reveals.
Ask SkimNews
