Anthropic gives six METR evaluators employee-like access to its systems and lets them publish risk findings without its approval

Anthropic has given six staff from METR, the AI safety nonprofit that tests frontier models, company laptops, office space, and internal system access roughly equal to what its own risk assessment teams have. METR can publish its findings on Anthropic’s risk levels, incidents, and practices, and Anthropic can’t veto publication. However, Anthropic can force redactions in set categories such as customer data, privileged legal material, and model architecture, and it decides what falls into those categories. METR is the first evaluator at the deepest of three access tiers Anthropic is building, while SecureBio and Faculty, Accenture’s AI unit, are working with narrower access and publication rights for now.