Anthropic says six METR employees now have employee-like access

On 15 October 2026, Anthropic published An update on independent evaluators. The page’s published time is 17:20 UTC, and its modified time is 23:35 UTC the same day. Anthropic says it has started to give employee-like access to a team from METR, which it describes as a nonprofit that evaluates frontier AI models, and that METR will publish the results of its risk assessments. TechCrunch, Aditya Mehta, timestamped 19:14 UTC, uses the phrase employee-level access for the same post. The wording below follows Anthropic unless a sentence is marked as TechCrunch’s.

What Anthropic says the six employees can see

Anthropic says it has onboarded six METR employees with company hardware, office space, and access to internal tools and systems “roughly on par with what our own risk assessment teams have access to.” TechCrunch reports that hardware as Anthropic-issued company laptops. The post says “company hardware.”

Outside a small number of categories, Anthropic says its employees are encouraged to share information with that team as they would with a fellow employee working on alignment risk. The categories it names are data about customers, attorney-client privileged information, and information about model architectures. A footnote says information about architectures presents a high risk of inadvertent leaks, and that its relevance to risk assessment is typically low. When it is highly relevant, Anthropic says it will generally seek to share it, though that will require explicit internal approvals. The footnote also says Anthropic exercises caution because these evaluators may work with other companies.

The engagement, Anthropic says, is limited to a small team under enforceable agreements. Other METR employees are siloed from that work, including researchers on other Anthropic projects. Anthropic says this gives a few qualified people, under its security standards, a close view of how it operates, and that it is not a route for other organizations to obtain Anthropic’s information or its customers’ and partners’ information.

Publication

Anthropic says METR will publish its analysis of Anthropic’s risk levels, incidents, or practices. The post says Anthropic cannot veto publication, and that Anthropic can require METR to redact information if it falls into narrow, predefined categories like the ones above. METR, it says, is free to characterize those redactions publicly and to say whether it believes the redactions were justified, as long as it does so without revealing the redacted information.

TechCrunch’s account of the publication terms matches those sentences. The TechCrunch story does not include a separate statement from METR. METR’s site, opened on 16 October, and METR’s 14 October proposal on how the public could be told about loss-of-control risks, do not state that these six employees have been onboarded. That 14 October post discusses evaluation access as something METR is proposing. It is not a confirmation of Thursday’s arrangement.

Three tiers, and the other names

Anthropic says it is developing three tiers of access and hopes, over time, to have several evaluators at the third tier, each publishing its own analysis.

In the first tier, evaluators conduct narrowly scoped risk investigations. Anthropic decides what access to provide, and evaluators can publish their findings with Anthropic’s approval.

In the second tier, evaluators can publish without Anthropic’s approval, and they have access to more information. Anthropic can still require removal of specific categories of information. Evaluators can publish their views on whether the redactions were justified.

The third tier, Anthropic says, is the arrangement it has described with METR: employee-like access, and publication rights similar to the second tier.

Anthropic also says its other partnerships are still in early stages, with more limited information and more limited publication rights. It names SecureBio and Accenture as evaluators it is working with, and says it is in early discussions with several other organizations, including Hugging Face.

On SecureBio, the post says SecureBio is reviewing the chemical and biological weapons chapter of Anthropic’s August Risk Report, after completing and publishing a pilot review of the previous risk report. After this review is complete, Anthropic says it will consider moving to projects with increased access. A footnote says the page was updated on 15 October to clarify SecureBio’s current project terms. TechCrunch describes SecureBio as an example of the bottom tier. Anthropic’s post does not give SecureBio a tier number.

With Accenture and Faculty, Anthropic says it is working toward a partnership that includes evaluating and red-teaming models, conducting alignment assessments, testing model safeguards, and building capabilities across the organizations. The post does not say that partnership is in place at the METR level of access, and it does not assign Faculty or Accenture a tier number.

Anthropic compares the idea to bank supervision and says the analogy is not exact: resident examiners work inside the largest banks because the bank knows more about its own risks than outsiders do. The post says Anthropic is embedding evaluators voluntarily. It does not present the arrangement as evidence that a model is safe.