Anthropic is granting outside safety researchers from the nonprofit METR employee-like access to evaluate its frontier artificial intelligence models. The embedded team can inspect internal systems and independently publish risk assessments without Anthropic holding veto power.
FireBrief
FireBrief