What happened
Anthropic announced it has given six employees from METR, a nonprofit that evaluates frontier AI models, "employee-like access." The team receives company hardware, office space and access to internal tools and systems roughly on par with Anthropic's own risk assessment teams.
Why it's doing this
According to Anthropic, as models become more powerful they require stronger safeguards and more challenging risk assessments. The company says it cannot publicly share all information relevant to its safety work, such as customer data, attorney-client privileged information or model architecture details. Instead, it gives third-party evaluators deep access and has them share evidence and reasoned conclusions about the risks.
Freedom to publish
METR will publish its analysis of Anthropic's risk levels, incidents or practices. Anthropic cannot veto publication; it can only require redactions in narrow, predefined categories. METR is free to characterize those redactions publicly.
Other evaluators
Anthropic said it is also working with SecureBio and Accenture as evaluators and is in early discussions with several other organizations, including Hugging Face. Last month the company committed to stronger independent, third-party assessment of its models and safety practices and signed the White House Accord on Super Intelligence.
