Anthropic gives six METR evaluators employee-like access and no veto over what they publish
Anthropic has given six staff from METR, a non-profit that tests AI models, company hardware, office space and internal tools roughly on par with its own risk teams, excluding customer data, privileged legal material and model architectures. METR can publish without Anthropic's approval, but Anthropic can demand redactions in predefined categories and decides what falls into them; the post's footnotes were revised twice within a day of publication.
Sources
Related
- METR: early-2025 AI slowed experienced developers by 19%
- Trump says AI leaders signed a self-policing "constitution"
- Anthropic tests whether AI could copy itself onto cheap household devices
- A PHP flaw stalled for two years was fixed ten weeks after AI-assisted researchers showed it could take over servers