Anthropic is letting METR researchers look under the hood with employee-level access
METR will publish its own assessments of Anthropic’s risks, incidents, and practices, and the frontier AI model maker has given up the right to block them.
TL;DR
- Anthropic is providing employee-level access to six METR researchers to independently assess its AI models, safety practices, and risk management.
- This arrangement is a response to recent AI agent incidents and increased calls for independent AI company checks.
- METR will have access to Anthropic's internal tools, office space, and company laptops, similar to Anthropic's own risk assessment teams.
- METR will publish its own assessments of Anthropic's risks, incidents, and practices, with Anthropic forfeiting the right to block publication.
- Anthropic can request redactions for specific categories like customer data or privileged information, but METR can publicly comment on these redactions.
- METR is the top tier of Anthropic's three-tier evaluation system, offering the most extensive access.
- This initiative aligns with a promise made by Anthropic CEO Dario Amodei and a White House accord signed by major AI companies, including OpenAI, Google, Meta, Nvidia, and xAI, committing to independent review.