Artificial intelligence firm Anthropic has provided the nonprofit organization METR with inside access to evaluate its AI models and internal safety protocols. According to a report by The Decoder, the evaluation initiative allows external researchers to examine how the company builds and monitors its systems.
The Role of External AI Audits
Independent evaluations by external safety groups have become an increasingly prominent mechanism for testing frontier artificial intelligence systems. Allowing third-party researchers inside facilities to inspect model behavior provides a different perspective than standard internal testing. However, the announcement did not specify which particular models METR is currently reviewing, nor did it disclose the exact scope or duration of the access granted.
What Remains Unclear
Aside from the lack of details regarding specific models under evaluation, the announcement did not clarify whether METR will publish a public report detailing its findings. Observers will be watching to see if future updates reveal the metrics used during these assessments or if the collaboration leads to permanent external oversight procedures at Anthropic.
As AI developers face mounting scrutiny over the safety of advanced systems, transparency around external testing methods remains limited.
Lead image: RDNE Stock project / Pexels
