Anthropic gives METR employee-like access
On 15 October 2026 Anthropic announced that it had started to give employee-like access to a team from METR, a non-profit that evaluates frontier AI models, and that METR will publish its own assessments of Anthropic's risks, incidents and practices. Six METR employees have been onboarded; Anthropic cannot veto publication but may require redaction in predefined categories. It is also working with SecureBio, Accenture and Faculty.
Why it matters
This is the step that Dario Amodei's essay of 12 September promised: outside evaluators inside the company with the right to publish. Anthropic compares it to resident examiners in banks and sets out three tiers of access for evaluators, of which METR's is the highest. How much METR will be able to say depends on terms that the page describes only in outline and that Anthropic edited on 15 and 16 October.
What the page says. "Last month" Anthropic committed to stronger independent third-party assessment of its models and safety practices and signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators. It has now started to give "employee-like access" to a team from METR. Six METR employees have been onboarded with company hardware, office space and access to internal tools and systems roughly on par with Anthropic's own risk assessment teams. Outside a small number of clearly defined categories (customer data, privileged legal information, information about model architectures) staff are encouraged to share information with them as with a colleague. METR will publish its own analysis of Anthropic's risk levels, incidents or practices; Anthropic cannot veto publication but can require redaction of information in predefined categories (for example reproducing confidential IP) and decides what falls into them; METR reserves the right to make statements about the level of access and information it received. The engagement is limited to a small team under enforceable agreements, and other METR employees are siloed from its work. Other evaluators. SecureBio is reviewing the chemical and biological weapons chapter of the August Risk Report after a pilot review of the previous one; with Accenture and Faculty Anthropic is working toward a partnership on evaluation and red-teaming, alignment assessments and safeguard testing; it is in early talks with several others, Hugging Face among them. The company is developing three tiers of access: in the first, evaluators run narrowly scoped investigations, Anthropic decides their access and they publish with its approval; in the second they can publish without approval and have more information, but Anthropic may require removal of specific categories and the evaluators may say publicly whether the removals were justified; the third is the arrangement with METR. Changes to the page. Footnote 2 reads "Updated on October 16 to clarify METR's current project terms with Anthropic"; footnote 3 "Updated on October 15 to clarify SecureBio's current project terms". So the text was changed after first publication, and the later version was read here. What the record does not claim. The terms of the agreement with METR were not read: only Anthropic's page describes them, and there is as yet no METR analysis under this agreement. The signing of the White House Accord is reported on the page's word; the document itself was not read here.