Anthropic has provided a team from the nonprofit METR inside access to audit its artificial intelligence models and safety practices. Six METR staffers received company hardware, office space, and access to internal tools, positioning them roughly on par with Anthropic's own risk teams. While customer data, attorney-client privileged information, and model architecture details remain off-limits, METR will independently publish its findings. Anthropic retains the right to request redactions in predefined categories, but METR maintains the freedom to publicly state whether it believes those redactions were justified.
This agreement surpasses the level of access Anthropic has previously offered to other outside reviewers. Organizations such as SecureBio, Accenture, and Faculty received more restricted access and tighter limits on publication, while Anthropic has also initiated early discussions with Hugging Face. The initiative follows Anthropic's signing of the White House Accord on Super Intelligence, which includes a formal commitment to collaborating with external evaluators on AI safety.
This partnership matters because it establishes a new transparency benchmark for AI safety evaluations through independent audits.