Anthropic says it has begun giving employee-like access to a team of six METR employees, who will publish their analysis of Anthropic's risk levels, incidents or practices without an Anthropic veto but subject to redaction of predefined categories that Anthropic decides. It says it is also working with SecureBio and Accenture as evaluators, and with Accenture and Faculty toward a broader partnership, is in early talks with others including Hugging Face, and is developing three tiers of access with different publication rights. The post is Anthropic's statement of its commitments and arrangements; it reports no evaluation results.
It sets out, in the developer's own words, what access and publication rights an embedded external evaluator would have, which determines how far its published conclusions could be checked.
Key facts
As stated in the sources, with where to find them.
- Anthropic says it onboarded six METR employees with company hardware, office space, and access to internal tools and systems roughly on par with its own risk assessment teams. Outside a few defined categories (customer data, attorney-client privileged information, model architecture information), staff are encouraged to share information with them as with a colleague. The engagement is limited to a small team under enforceable agreements, with other METR staff siloed from its work.What we mean by employee-like access
- Anthropic says METR will publish its analysis of Anthropic's risk levels, incidents or practices; Anthropic cannot veto publication but can require redaction of information in predefined categories, and Anthropic decides what falls into them. METR reserves the right to make certain statements about the access and information it received. A footnote says the page was updated on 2026-10-16 to clarify METR's project terms.What we mean by employee-like access; footnote 2
- Anthropic lists four things evaluators need: expertise in assessing frontier models, independence (no equity in, or financial or operational dependence on, the company), enough access to evaluate risks including training and safeguards, and freedom to publish. It offers bank supervision with resident examiners as a similar logic and says the analogy is not exact.Why we work with independent evaluators
- Anthropic describes three access tiers: in the first, evaluators run narrowly scoped investigations, Anthropic decides their access, and publication needs its approval; in the second, evaluators publish without approval, have more information, and may comment on whether redactions were justified; the third is the METR arrangement, with employee-like access and second-tier publication rights.Our other evaluator partnerships
- Anthropic says SecureBio is reviewing the chemical and biological weapons chapter of its August Risk Report after a published pilot review of the previous report, and that it will consider projects with more access afterward. It describes an ambitious partnership in progress with Accenture and Faculty covering model evaluation, red-teaming, alignment assessments and safeguard testing, and early discussions with other organizations including Hugging Face.Our other evaluator partnerships; footnote 3
- Anthropic says that in September 2026 it committed to stronger independent assessment, and that it signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators. It says it is embedding evaluators voluntarily and believes every frontier AI company should do the same. These are statements of intent and arrangements, not results of any assessment.Opening; closing section
- The post gives no scope, duration or publication timeline for METR's work. METR's own descriptions of the terms it seeks are in its October 14, 2026 posts.Whole post
Findings that cite this record
No tracked finding cites this record yet.