Anthropic is granting employee-level access to six researchers from METR, a nonprofit organization dedicated to testing frontier AI models, allowing them to independently evaluate Anthropic’s models, safety protocols, and risk management strategies. The company announced this significant development in a blog post on Thursday.
This initiative arrives amidst a period where reports of AI agents acting outside their intended instructions have surfaced nearly every week or two since late August, intensifying calls for independent oversight within AI firms.
“We want to be publicly accountable for the decisions we’re making,” Anthropic stated in the blog post. “We aim to close the gap between what evaluators understand about a frontier AI company’s systems and what the company itself understands.”
Under the new agreement, the METR team has been provided with Anthropic-issued company laptops, office space, and access to internal tools comparable to those used by Anthropic’s own risk assessment teams. “Anthropic employees are encouraged to share information with them just as they would with a fellow employee working on alignment risk,” the company explained.
Anthropic has drawn a parallel to the banking sector, comparing this arrangement to how regulators place resident examiners inside the largest financial institutions.
METR will be responsible for publishing its own assessments regarding Anthropic’s risks, incidents, and practices, with the company relinquishing its ability to block these publications.
“Anthropic can’t veto publication, but we can require them to redact information if it falls into narrow, pre-defined categories,” the company noted, such as customer data, attorney-client privileged material, or information concerning model architectures. “METR is free to characterize our redactions publicly, and say whether it believes the redactions were justified, as long as it does so without revealing the redacted information,” Anthropic added.
Currently, METR is the sole evaluator at the highest level of a three-tier system Anthropic is constructing. Evaluators in the lowest tier, like the biosecurity nonprofit SecureBio, conduct narrow reviews and must obtain Anthropic’s approval to publish findings. The middle tier allows for more information and publication without approval, whereas the top tier, occupied by METR, grants employee-level access.
Thursday’s announcement fulfills a commitment made by Anthropic CEO Dario Amodei in a September essay, in which he proposed embedding outside evaluators inside AI laboratories as a preliminary measure to slow the technology’s advancement. OpenAI has since pledged a similar approach, and several major tech entities, including Google, Meta, Nvidia, and xAI, have signed a White House accord pledging their commitment to independent reviews.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.