Anthropic said on October 15, 2026 that it has started giving a team from METR, a nonprofit that evaluates frontier AI models, “employee-like access” to the company. In its announcement, Anthropic says six METR employees have been onboarded with company hardware, office space and access to internal tools and systems “roughly on par” with what its own risk assessment teams use, and that METR will publish the results of its risk assessments without an Anthropic veto.
The move follows Anthropic’s September commitment to stronger independent, third-party assessment of its models and safety practices, and its signature on the White House Accord on Super Intelligence, in which companies pledged to work with independent external auditors or evaluators. Anthropic says it has worked with METR over the past year on several reviews of its Risk Reports.
On this page
What METR can see and publish
| Item | Terms, per Anthropic |
|---|---|
| Team | Six METR employees, under enforceable agreements |
| Access | Company hardware, office space, internal tools and systems roughly on par with Anthropic’s risk assessment teams |
| Excluded information | Customer data, attorney-client privileged information, information about model architectures, and a small number of other defined categories |
| Publication | METR publishes its analysis of Anthropic’s risk levels, incidents or practices; Anthropic cannot veto it |
| Redactions | Anthropic can require redactions only in narrow, predefined categories; METR may say publicly whether it thinks a redaction was justified |
| Separation | Other METR employees, including researchers working with Anthropic on other projects, are siloed from the embedded team |
Anthropic explains the exclusion of architecture information in a footnote: because evaluators may also work with other companies, it treats architecture details as a high risk of inadvertent leaks and usually of low relevance to risk assessment. When such information is highly relevant, the company says it will generally seek to share it with explicit internal approval.
The company compares the arrangement, with the caveat that the analogy is not exact, to bank supervision, where regulators place resident examiners inside the largest banks because the banks know more about their own risk positions than outsiders do.
Three tiers of evaluator access
Anthropic says it is building three tiers of access for outside evaluators:
- First tier: narrowly scoped risk investigations. Anthropic decides what access to provide, and evaluators publish their findings with Anthropic’s approval.
- Second tier: more information and publication without Anthropic’s approval. Anthropic can still require removal of specific categories of information, and evaluators can publish their views on whether those redactions were justified.
- Third tier: the METR arrangement, with employee-like access and publication rights similar to the second tier.
METR is the only evaluator Anthropic describes at the third tier. The company says it hopes to have several evaluators there over time, each publishing its own analysis of Anthropic’s risk posture and key decisions.
SecureBio, Accenture and Hugging Face
Anthropic names three other organisations in the post. 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 report; Anthropic says it will consider projects with more access after this review. With Accenture and its AI business Faculty, Anthropic says it is working toward a partnership that covers evaluating and red-teaming models, alignment assessments and testing model safeguards, the arrangement it first announced on September 18. Anthropic also says it is in early discussions with several other organisations, including Hugging Face.
The post was updated on October 15 to clarify the terms of SecureBio’s current project with Anthropic.
What changes
In September, the Accenture partnership described embedded evaluation as a plan. METR is now the only outside team Anthropic places at the third tier: already working inside the company, with access close to its own risk teams and the right to publish without company approval. Anthropic says it is embedding evaluators voluntarily and believes every frontier AI company should do the same; the post does not say when METR’s first assessment will be published.
Earlier this month Anthropic also set out three access tiers for its Cyber Verification Program, a separate scheme for security researchers. Anthropic’s ownership and company facts are on its AI companies profile, and its model releases are listed in AI model updates.
