Anthropic said on 15 October 2026 that it has started giving “employee-like access” to a team from METR, a nonprofit that evaluates frontier AI models, and that it cannot veto what METR publishes about Anthropic’s risk levels, incidents or practices.
The company framed the move as following through on a commitment to stronger independent assessment and on the White House Accord on Super Intelligence, under which firms pledged to partner with independent external auditors or evaluators.
What “employee-like access” means
“We’ve onboarded six METR employees with company hardware, office space, and access to internal tools and systems that are roughly on par with what our own risk assessment teams have access to,” Anthropic wrote.
Outside a small set of clearly defined categories — for example data about customers, attorney-client-privileged information, or information about model architectures — Anthropic says its staff are encouraged to share information with the METR team as they would with a colleague working on alignment risk.
The engagement is limited to that small team under enforceable agreements. Other METR employees are siloed from that team’s work, including researchers who work with Anthropic on other projects, the post says.
Publication: no veto, narrow redactions
“METR will publish its analysis of our risk levels, incidents, or practices,” Anthropic said. “Anthropic can’t veto publication, but we can require them to redact information if it falls into narrow, pre-defined categories like the ones above.”
METR “is free to characterize our redactions publicly, and say whether it believes the redactions were justified, as long as it does so without revealing the redacted information,” according to the post.
That is Anthropic’s description of the METR contract. It is not an independent audit of how those rules will play out in practice.
Three tiers of evaluator access
Anthropic said it is developing three tiers and hopes, over time, to have several evaluators at the third tier, each publishing their own analysis.
Tier 1: Narrowly scoped risk investigations. Anthropic decides what access to provide, and evaluators can publish “with our approval.”
Tier 2: Evaluators can publish without Anthropic’s approval and get more information. Anthropic can still require removal of specific categories of information; evaluators can publish whether they think those redactions were justified.
Tier 3: The METR arrangement — employee-like access, with publication rights similar to tier 2 (no veto; narrow redactions).
SecureBio is currently in the first tier, reviewing the chemical and biological weapons chapter of Anthropic’s August Risk Report, after publishing a pilot review of an earlier report. Anthropic said that after this review it will consider moving to projects with more access and publication rights.
Anthropic also said it is working with Accenture and Faculty toward a partnership that includes evaluating and red-teaming models, alignment assessments and testing safeguards, and that it is in “early discussions with several other organizations, including Hugging Face.”
Contrast with earlier Reuters reporting
Two days earlier, Reuters reported — covered on TSN as post 21436 — that Anthropic and OpenAI had sought significant control over what outside evaluators can publish, citing two people familiar with the discussions. Anthropic told Reuters that seeking significant control over embedded evaluators’ publications “did not accurately characterize all of its discussions,” and that with evaluators it had worked with before it had made clear the reviewers control methods, criteria, conclusions and what to publish, while for new firms it reserved “the right to comment on and approve reports” while they learn the models.
Today’s Anthropic post is the company’s own account of the METR deal: no publication veto, only narrow redactions, and METR able to say publicly whether redactions were justified. That is a clearer public commitment on veto power than Reuters’ sources described for the broader talks. It does not by itself settle what Anthropic sought in earlier negotiations, which remains a contested Reuters report.
Why Anthropic says it matters
Anthropic argued that frontier models need stronger safeguards and harder risk assessments, but that companies cannot publish everything relevant to safety work because some of it could help malicious actors or is commercially sensitive. Embedding evaluators, it said, aims to “close the gap between what evaluators understand about a frontier AI company’s systems and what the company itself understands.”
It compared the idea, loosely, to resident bank examiners. “We’re embedding evaluators voluntarily, and we believe every frontier AI company should do the same.”
What we don’t know
- When METR’s first public analysis under this arrangement will appear.
- How often Anthropic will invoke redactions, and whether METR will dispute them.
- Whether Accenture, Faculty or Hugging Face will reach tier 3, and on what timeline.
What this means
- METR inside the building. Six staff get access Anthropic likens to its own risk teams.
- No veto on METR’s analysis. Anthropic says it can only force narrow redactions, which METR may criticise publicly.
- A ladder for other evaluators. SecureBio starts at tier 1 approval-gated publication; Anthropic says it wants more teams at the METR tier.
Sources
- Anthropic, “An update on independent evaluators”, 15 October 2026
- TSN, “Anthropic and OpenAI Sought Significant Control Over What Outside AI Evaluators Can Publish, Reuters Reports”
Related stories
- Anthropic and OpenAI Sought Significant Control Over What Outside AI Evaluators Can Publish, Reuters Reports
- METR Offers Frontier AI Labs Independent, Public Assessments of Loss-of-Control Risk, but Says They Are Not Audits
- FTC Sends Civil Investigative Demands to OpenAI and Anthropic Over AI Safety Incidents, Axios Reports
Follow @tsnmediahub on X for breaking AI news as it happens.
Follow on X
