Anthropic is taking a notable step toward greater transparency in frontier AI by allowing researchers from METR to examine its systems and practices with employee-level access.
METR will be able to publish its own assessments of Anthropic’s risks, incidents, and safety processes. Importantly, Anthropic has reportedly given up the right to block those publications, giving the review process a stronger measure of independence.
Why this matters
Independent evaluation is one of the most important tools for building safer AI. By opening its doors to outside experts, Anthropic is helping create a model for how advanced AI companies can be scrutinized without relying solely on internal claims.
- More transparency around frontier AI risks
- Stronger accountability for safety practices
- Potential industry precedent for independent audits
While the impact will depend on the depth and regularity of METR’s assessments, this is a promising win for AI governance: real access, independent reporting, and a clearer path toward trustworthy AI development.