Anthropic gives METR employee-like access, and METR can publish without a veto

On October 15, Anthropic said it has started giving employee-like access to a team from METR. The company's post on X points at that write-up. Six METR employees are onboarded with company hardware, office space, and internal tools "roughly on par" with Anthropic's own risk-assessment teams.

The exclusions are named: customer data, attorney-client privileged information, and model-architecture details. Anthropic's privacy policy is the customer-data rule those evaluators do not get past. METR will publish its analysis of risk levels, incidents, or practices. Anthropic cannot veto publication. It can require redaction inside narrow predefined categories, and METR can say in public whether it thinks a redaction was justified, without revealing the redacted material.

Anthropic describes three access tiers. The first is a narrow investigation that Anthropic approves before publication. The second can publish without approval, with the same kind of category redactions. METR is the third tier: the employee-like access above, with publication rights like the second tier. SecureBio is reviewing the chemical and biological weapons chapter of the August risk report, as a more limited engagement. Anthropic is also working with Accenture and Faculty, and is in early discussions with other organizations, including Hugging Face.

CNBC reports that Anthropic says it is not funding METR's work. Reuters, via Investing.com, and TechCrunch reported the same access arrangement.

What it costs

Nothing on a Claude plan changes. This is an evaluator setup, not a price.

Should you switch

No. It does not change the product you use. It landed the day after the reported FTC investigative demands to OpenAI and Anthropic. It is the most concrete third-party audit arrangement a frontier lab has published. The thing to watch is METR's first report, and whether any other lab matches the no-veto term.