METR on AI labs should give independent evaluat…

METR

Nonprofit AI safety research organization that evaluates frontier AI systems.

Starting in February 2026, METR conducted a pilot exercise to assess misalignment risks from AI agents used inside frontier AI developers, with participation from Anthropic, Google, Meta, and OpenAI. Each participant provided access to their most capable internal model(s), including raw chains of thought, and a wide range of non-public information about the model(s), how AI was used and monitored internally, and trends in the pace of progress. We believe that periodic third-party assessment of risks from developers’ internal use of AI should be adopted throughout the industry.

AI Verified (May 19, 2026)

Source:

METR, “Frontier Risk Report (February to March 2026),” May 19, 2026.

20d ago

Policy proposals and claims

AI labs should give independent evaluators employee-like access

votes For

Statement relation verification history AI Verified Report this

AI Verified METR describes independent third-party access to internal frontier models, raw chains of thought, and broad non-public operational information, then recommends periodic adoption—employee-like evaluator access. · Hector Perez Arenas gpt-5 · 19d ago

Vote inference verification history AI Verified Report this

AI Verified METR explicitly recommends periodic third-party assessment after describing unusually broad internal access; this supports independent evaluators receiving employee-like access. · Hector Perez Arenas gpt-5 · 19d ago

Quote authenticity verification history

Report this

AI Verified METR's May 19, 2026 Frontier Risk Report contains this account and its conclusion verbatim; METR is the publishing organization. · Hector Perez Arenas gpt-5 · 19d ago

replying to METR

Other opinions from METR

See all

METR votes For

AI labs should conduct risk assessments before high-stakes internal deployment of frontier models

Most importantly, third-party evaluations triggered by public deployments are not designed to cover risks from internal use within the developer itself. [...] We believe that periodic third-party assessment of risks from developers’ internal use of A...

more AI Verified source (May 19, 2026)