Anthropic gives METR deeper access to internal systems for independent risk checks

METR can now assess Anthropic’s model and safety risk with access closer to the company’s own teams, and its findings will be published.

Anthropic said it has begun giving METR employee-like access to its internal systems so the nonprofit can independently assess model and safety risk, a move that matters because it gives outside evaluators deeper visibility into frontier AI development. The company said six METR employees now have company hardware, office space, and access comparable to Anthropic’s own risk teams, while METR can publish its findings and Anthropic cannot block publication, though it can require narrow redactions. Anthropic also said METR will publish its risk assessments, and that it is already working with SecureBio and Accenture while discussing possible future partnerships with other groups such as Hugging Face. The post frames the arrangement as part of a broader three-tier evaluator model, with SecureBio currently in the first tier and the goal of eventually embedding more independent evaluators.