Here is an update on independent evaluators.

Anthropic has begun giving a team from METR, a nonprofit that evaluates frontier AI models, access to its internal systems roughly equivalent to that of its own employees. The arrangement, announced on October 15, 2026, marks a concrete step in the company's pledge to strengthen independent, third-party assessment of its models and safety practices. METR will publish the results of its risk assessments, and Anthropic cannot veto publication.

The move follows two earlier commitments. Last month, Anthropic said it would pursue stronger independent evaluation. It also signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators. Over the past year, Anthropic has worked with METR on several reviews of its Risk Reports. It is also working with SecureBio and Accenture as evaluators, and is in early discussions with several other organizations, including Hugging Face.

Why independent evaluators need deep access

As models have grown more powerful, they have required stronger safeguards and more challenging risk assessments. Anthropic argues it cannot publicly share all information relevant to its safety work, because some of it could help malicious actors or is commercially sensitive, such as compute or training information. Instead, the company is giving third-party evaluators deep access and having them publish evidence and reasoned conclusions about the risks involved in building its models. The stated goal is to close the gap between what evaluators understand about a frontier AI company's systems and what the company itself understands.

Anthropic lists four requirements for effective evaluation. Evaluators need expertise in assessing frontier model capabilities, including understanding how evaluations can fail, such as alignment faking. They need independence: no equity in the company being evaluated and no financial or operational dependence on it. They need enough access to meaningfully evaluate risks, including visibility into how models were trained and how safeguards function. And they need freedom to publish observations and assessments.

The company compares the logic to bank supervision. Banks know more about their own risk positions than outsiders, so regulators assign resident examiners to work inside the largest banks. If a bank's risk-taking could threaten the public, third parties try to see that in advance.

What employee-like access means in practice

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 use. Outside a small number of clearly defined categories—customer data, attorney-client privileged information, and information about model architectures—Anthropic employees are encouraged to share information with them as they would with a colleague working on alignment risk.

METR will publish its analysis of Anthropic's risk levels, incidents, or practices. Anthropic cannot veto publication, but it can require redaction of information falling into narrow, predefined categories. METR is free to characterize those redactions publicly and say whether it believes they were justified, without revealing the redacted information.

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. Anthropic frames embedded evaluation as a close view for a few qualified people working under its security standards, not a channel for other organizations to obtain its information or that of its customers and partners.

A three-tier access model

Anthropic is developing three tiers of access for evaluators, with the goal of eventually having several evaluators at the third tier, each publishing their own analysis of its risk posture and key decisions.

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

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

The third tier is the arrangement with METR: employee-like access and publication rights similar to the second tier.

Other partnerships and what comes next

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. Once that review is complete, Anthropic will consider moving to projects with increased access. With Accenture and Faculty, Anthropic is working toward a partnership that includes evaluating and red-teaming models, conducting alignment assessments, testing model safeguards, and building capabilities across the organizations.

Because these other partnerships are still early, those evaluators receive more limited information and have more limited publication rights. Anthropic says the goal at this stage is to ensure partners protect confidential information and have the resources and expertise to evaluate its models, with fuller embedding to follow as trust builds.

Anthropic notes one caution: because evaluators, like bank examiners, may also work with other companies, there is a risk of inadvertent leaks. Information about architectures presents a high risk of inadvertent leaks and is typically of low relevance to risk assessment. When it is highly relevant, Anthropic says it will generally seek to share it, subject to explicit internal approvals.

The company says it is embedding evaluators voluntarily and believes every frontier AI company should do the same. For readers, the key question to watch is whether METR's published assessments prove substantive and whether Anthropic expands third-tier access to additional evaluators, turning a voluntary arrangement into a broader norm for frontier AI oversight.