METR Defends Anthropic Evaluation Deal as Better Than Self-Assessment

METR's Seán Ó hÉigeartaigh explained why the organization accepted Anthropic's evaluation terms: the alternative was evaluating nothing at all or relying on Anthropic's own unsupported self-assessments. Imperfect third-party evaluation, he argued, still beats no external checks.

2026-11-07 ~ 2026-11-07 · 2 related posts

• METR: getting nothing would be worse than accepting Anthropic's eval terms — S_OhEigeartaigh · 2026-11-07
• METR debate: third-party evals beat Anthropic's unevidenced self-estimates, argue researchers — S_OhEigeartaigh · 2026-11-07