Anthropic Grants METR Embedded 'Employee-Like' Access for Third-Party Model Risk Evaluation
Anthropic has given external evaluators 'employee-like access' to assess its models' risks, following through on a commitment to strengthen safeguards. METR, a nonprofit that has evaluated models at OpenAI...