On 15 October 2026, Anthropic updated its independent-evaluator arrangements: it has started giving a team from the nonprofit METR “employee-like” access to assess its models and safety practices more deeply. The company says that last month it committed to stronger independent third-party assessment and signed the White House Accord on Super Intelligence, under which companies pledged to partner with independent external auditors or evaluators.
Why independent evaluators
Anthropic notes that as models grow more powerful, they need stronger safeguards and more demanding risk assessments. It wants to be publicly accountable for its decisions, but cannot publish everything relevant to its safety work—some of it could help malicious actors, or is commercially sensitive, such as compute or training information. Instead, it is giving third-party evaluators deep access and having them share evidence and reasoned conclusions, aiming to close the gap between what evaluators understand about a frontier AI company’s systems and what the company itself understands.
Evaluators of frontier models, Anthropic says, need expertise in assessing capabilities and failure modes (including alignment faking); independence (no equity in the company evaluated, and no financial or operational dependence on it); enough access to evaluate risks meaningfully—beyond the latest models, into how they were trained and how safeguards work; and freedom to publish observations and assessments. The post draws an analogy to bank supervision and resident examiners; the analogy is illustrative, not a legal equivalence.
What “employee-like access” means
Anthropic has onboarded six METR employees with company hardware, office space, and access to internal tools and systems roughly on par with what its own risk-assessment teams have. Outside a small number of clearly defined categories—for example customer data, attorney-client privileged information, or information about model architectures—Anthropic employees are encouraged to share information with them as they would with a fellow employee working on alignment risk.
METR will publish its analysis of Anthropic’s risk levels, incidents, or practices. Anthropic cannot veto publication, but can require redaction of information that falls into predefined categories (Anthropic ultimately decides what falls into those categories; METR reserves the right to make certain statements about the level of access and information it received). A footnote updated on 16 October clarifies METR’s current project terms with Anthropic.
The engagement is limited to a small team operating under enforceable agreements. Other METR employees are siloed from that team’s work, including researchers who work with Anthropic on other projects. Embedded evaluation gives a few qualified people working under Anthropic’s security standards a close view of how the company operates—not a channel for other organizations to obtain Anthropic’s information, or that of its customers and partners.
Figure: Left — internal risk-review collaboration (illustration); right — an independent evaluator publishing under glass-box constraints — editorial illustration with no text or logos; not an actual office or system photo.
Other partnerships and a three-tier model
Anthropic is also working with SecureBio and Accenture, and is in early discussions with several other organizations, including Hugging Face. Because those other partnerships are still early, evaluators currently receive more limited information and more limited publication rights; as trust builds, Anthropic hopes to fully embed more independent evaluators.
It is developing three tiers of access:
- First tier: Narrowly scoped risk investigations; Anthropic decides what access to provide, and evaluators publish with Anthropic’s approval.
- Second tier: More information, and evaluators can publish without Anthropic’s approval; Anthropic can still require removal of predefined categories of information, and evaluators can publish their views on whether redactions were justified.
- Third tier: The METR arrangement—employee-like access, with publication rights similar to the second tier.
SecureBio is reviewing the chemical and biological weapons chapter of Anthropic’s August Risk Report, after completing and publishing a pilot review of a previous risk report; after that review, Anthropic will consider projects with increased access. With Accenture and Faculty, it is working toward a partnership that includes evaluating and red-teaming models, alignment assessments, testing safeguards, and building capabilities across organizations. Anthropic says it is embedding evaluators voluntarily and believes every frontier AI company should do the same.
Over the past year, Anthropic has worked with METR on several reviews of its Risk Reports; this announcement marks a move from report review toward deeper embedded evaluation.
Primary source:
- Anthropic, “An update on independent evaluators”, 15 October 2026 (footnotes updated 15–16 October), https://www.anthropic.com/news/an-update-on-independent-evaluators
