Anthropic gives METR employee-like access to audit its frontier AI models

An update on independent evaluators

Anthropic gives METR employee-like access to audit its frontier AI models

Anthropic has granted METR, a nonprofit evaluator, employee-like access to its internal systems, hardware, and tools to assess risks in its frontier models. METR can publish its findings, with Anthropic able to redact only narrow, predefined categories. The company is also working with SecureBio and Accenture, and has outlined a three-tier access system for independent evaluators.

We're embedding evaluators voluntarily, and we believe every frontier AI company should do the same.

More from this day

2026-10-15