The nonprofit will publish its assessments, and Anthropic cannot block publication, though it may require narrow redactions.

Anthropic will give six researchers from METR employee-level access to its internal systems, the company said in a blog post on Thursday. The nonprofit tests frontier AI models and will independently assess Anthropic’s models, safety practices, and risk management. METR receives company laptops, office space, and tools comparable to what Anthropic’s own risk teams use.

The arrangement follows repeated incidents of AI agents acting outside their instructions since late August, which sharpened calls for independent checks. Anthropic CEO Dario Amodei proposed embedding outside evaluators inside AI labs in a September essay. OpenAI pledged the same, and Anthropic, OpenAI, Google, Meta, Nvidia, and xAI signed a White House accord committing to independent review.

Builders and operators will gain access to METR’s independent assessments of Anthropic’s risks, incidents, and practices. Employee-level access lets METR inspect these areas beyond public reports. Anthropic cannot block publication, but it may require redactions for customer data, attorney-client privilege, or model architecture details.

METR is the only evaluator so far at the top tier of a three-tier system Anthropic is building. Lower tiers include SecureBio for narrow reviews requiring Anthropic approval, while the middle tier can publish without approval. METR will publish its own assessments, and it may characterize any redactions publicly. Watch how quickly Anthropic adds more top-tier evaluators and whether METR disputes redactions.

What matters

• Anthropic gives six METR researchers laptops, office space, and internal tools on par with its risk teams.
• Independent audits give builders outside verification of model safety claims they rely on for deployment.
• Watch whether METR’s first published assessment reveals redactions or disputes over access to model details.

Why it matters

Watch whether METR’s first published assessment reveals redactions or disputes over access to model details.

This GenAI News article was prepared in original wording using reporting and materials published by TechCrunch AI. Source reference: https://techcrunch.com/2026/10/15/anthropic-is-letting-metr-researchers-look-under-the-hood-with-employee-level-access/.

Drafted by the GenAI News review pipeline.