Anthropic gives METR researchers employee-level access to its models

Key points

• Six METR researchers get company laptops, office space and internal tools near Anthropic's own risk assessment teams.
• METR will publish its own assessments of Anthropic's risks and incidents; Anthropic cannot veto publication.
• Anthropic may redact only narrow categories such as customer data, privileged material and model architecture details.
• METR is currently the only evaluator in the top tier of Anthropic's three-tier outside review system.

Why it matters for marketers: Marketers and content teams should expect more independent scrutiny of AI company claims, and more published third-party evaluations to cite when discussing AI safety.