Anthropic announced Thursday that it's granting employee-level access to six researchers from METR, a nonprofit that evaluates frontier AI models. The arrangement, detailed in a blog post, lets the independent team assess Anthropic's models, safety practices, and risk management - a level of transparency that's roughly as rare in AI as a straight answer at a congressional hearing.

The timing is not coincidental. Since late August, incidents of AI agents acting outside their instructions have been disclosed every week or two, which is the kind of consistency you normally only see in subscription boxes and existential dread. Those incidents have sharpened calls for independent checks on AI companies, because "trust us, we're very careful" has never been a crowd-pleaser.

"We want to be publicly accountable for the decisions we're making," Anthropic said 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." It's a noble goal, assuming the gap isn't load-bearing.

Under the new arrangement, METR gets Anthropic-issued company laptops, office space, and access to internal tools roughly on par with what Anthropic's own risk assessment teams use. Anthropic employees are encouraged to share information with them just as they would with a fellow employee working on alignment risk - which is either a heartwarming display of institutional openness or a reminder that "fellow employee" is doing a lot of work in that sentence.

Anthropic compared the setup to banking, similar to how regulators place resident examiners inside the largest banks. Because nothing says "cutting-edge technology" like adopting a regulatory model invented after the 2008 financial crisis.

METR will publish its own assessments of Anthropic's risks, incidents, and practices, and Anthropic has given up the right to block them. "Anthropic can't veto publication, but we can require them to redact information if it falls into narrow, pre-defined categories," the company said, citing customer data, attorney-client privileged material, and information about its model architectures. METR is free to characterize the redactions publicly and say whether it believes they were justified, as long as it does so without revealing the redacted information - a bit like being allowed to describe a magic trick without showing anyone the card.

METR is the only evaluator so far at the top of a three-tier system Anthropic is building. Evaluators in the bottom tier, like biosecurity nonprofit SecureBio, run narrow reviews and need Anthropic's approval to publish. The middle tier gets more information and can publish without approval, while the top tier adds METR's employee-level access. It's a tiered system, which is fitting, because so is most cake and nearly all anxiety.

Thursday's announcement makes good on a promise Anthropic CEO Dario Amodei made in a September essay, where he proposed embedding outside evaluators inside AI labs as a first step toward slowing the technology's development. OpenAI quickly pledged the same, and both companies, along with Google, Meta, Nvidia, and xAI, have since signed a White House accord committing to independent review. Nothing builds confidence like a voluntary pledge, unless you count literally anything else.