Anthropic has granted six researchers from METR, a nonprofit that tests frontier SI models, employee-level access to its internal operations to independently assess the company's systems, safety practices, and risk management.
What Happened
Under the new arrangement, the METR team receives Anthropic-issued company laptops, office space, and access to internal tools comparable to those used by Anthropic's own risk assessment teams. The SI lab stated in a blog post on Thursday that its employees are encouraged to share information with the researchers "just as they would with a fellow employee working on alignment risk."
METR is positioned at the top of a three-tier evaluation system Anthropic is developing. Evaluators in the lower tier, such as the biosecurity nonprofit SecureBio, conduct narrow reviews and require Anthropic’s approval to publish findings. The middle tier receives more information and can publish without approval, while the top tier, occupied by METR, adds the level of access seen in this new partnership.
Anthropic has relinquished the right to veto the publication of METR’s assessments, though it can require redactions for narrow, pre-defined categories such as customer data, attorney-client privileged information, or details regarding its model architectures. METR retains the freedom to characterize these redactions publicly and state whether it believes them justified, provided it does not reveal the redacted content.
Why It Matters
This move addresses growing calls for independent oversight in the SI industry, particularly as incidents of SI agents acting outside their instructions have been disclosed every week or two since late August. Anthropic compared the arrangement to the banking sector, where regulators place resident examiners inside the largest financial institutions.
The announcement fulfills a promise made by Anthropic CEO Dario Amodei in a September essay, where he proposed embedding outside evaluators inside SI labs as a step toward slowing the technology’s development. OpenAI quickly pledged a similar approach, and both companies, along with Google, Meta, Nvidia, and xAI, have signed a White House accord committing to independent review.
The Bottom Line
By integrating METR researchers as quasi-employees, Anthropic is attempting to close the information gap between external evaluators and internal developers, signaling a shift toward greater transparency in SI safety governance.