The nonprofit can use company tools and publish assessments, with limited redactions allowed.
Anthropic has announced an arrangement giving six METR researchers employee-level access to its models, safety practices and risk management. METR is a nonprofit that tests advanced AI models.
According to TechCrunch, the researchers will receive Anthropic-issued laptops, office space and access to internal tools roughly comparable to those used by Anthropic’s own risk-assessment teams. Anthropic employees are encouraged to share information with them as they would with colleagues working on alignment risk, a term for reducing harmful or unexpected AI behavior.
That setup could give METR a closer look at how Anthropic evaluates risk and handles incidents. It does not yet show whether the access will produce stronger public findings; that will depend on what METR examines and reports.
METR will publish its own assessments of Anthropic’s risks, incidents and practices, TechCrunch reports. Anthropic has given up the right to block publication, but it can require narrow redactions for customer data, attorney-client-privileged material and information about model architectures.
METR is currently the only evaluator in the top tier of Anthropic’s three-tier review system. TechCrunch says lower tiers involve narrower reviews, while the top tier adds employee-level access.
The arrangement is an announced review framework, not an independent finding about Anthropic’s safety. Its practical value will become clearer when METR publishes its first assessment and explains what it examined, what it found and which details were withheld.