Anthropic seats six METR staff on its own hardware, with a redaction right but no publication veto
Anthropic has given six METR staff employee-like access to its internal tools and systems. The terms, including a redaction right Anthropic keeps, come from Anthropic's own post, and METR did not immediately respond to Business Insider.
The Board Room · Leadership desk
Six METR staff have company hardware, office space and internal tools roughly on par with Anthropic's risk teams. Anthropic cannot veto their publication, only require narrow redactions. Other METR staff are siloed. Early partners get more limited information and publication rights.
- What's now possible capability Embedded METR team of six Company hardware, office space and internal tools roughly on par with Anthropic's own risk teams, claim 3
- What it limits constraint Anthropic Cannot veto publication of METR's analysis, but can require redaction in narrow, predefined categories, claim 5
- What it limits constraint Other METR employees Siloed from the embedded team's work under enforceable agreements, claim 15
- Who's exposed exposure Early-stage evaluator partners Receive more limited information and have more limited publication rights, claim 7
| Who | How | Kind | Claim |
|---|---|---|---|
| Embedded METR team of six | Company hardware, office space and internal tools roughly on par with Anthropic's own risk teams | capability | 3 |
| Anthropic | Cannot veto publication of METR's analysis, but can require redaction in narrow, predefined categories | constraint | 5 |
| Other METR employees | Siloed from the embedded team's work under enforceable agreements | constraint | 15 |
| Early-stage evaluator partners | Receive more limited information and have more limited publication rights | exposure | 7 |
What happened
- Anthropic also named SecureBio and Accenture as evaluators it is working with, and said it is in early talks with several others, including Hugging Face.
- Those other partnerships are early, so their evaluators receive more limited information and have more limited publication rights than the METR team.
- Anthropic ties the step to its signing of the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators.
- Business Insider reported that the disclosure is the first time a frontier AI company has embedded evaluators in this way.
- Anthropic said it has worked with METR over the past year on several reviews of its Risk Reports.
Why it matters
- What it limits constraint Information about model architectures is outside the default sharing norm, which limits how far into the models the METR team is encouraged to look.
- What doesn't add up contradiction Business Insider describes free rein to publish with a couple of caveats, while Anthropic introduces its redaction categories as examples. Neither account settles how broad the categories are.
- Who's exposed exposure Anthropic's risk levels, incidents and practices become subject to outside analysis that it cannot block, only redact within defined categories.
- What it sets up precedent OpenAI has said it would embed evaluators but has not. Until it does, Anthropic's no-veto, redact-only terms are the only reference for what a frontier lab has offered.
Next story · Build · yesterday
Anthropic plans to put outside evaluators inside its labs, starting with AccentureConfirmed4 publishers
Anthropic presents employee-like access for a METR team [1] as a substitute for disclosure it says it cannot make. The company said it cannot publicly share all the information relevant to its safety work, because some of it could be useful to malicious actors or is commercially sensitive, such as compute or training information [18]. Anthropic said: "We aim to close the gap between what evaluators understand about a frontier AI company's systems and what the company itself understands." [16]
The model Anthropic reaches for is bank supervision, where regulators assign resident examiners to work inside the largest banks. The company itself says the analogy is not exact [14]. One difference is who does the assigning. A supervisor appoints its examiners, and Business Insider reported that Anthropic chose the METR evaluators [22].
Anthropic's own independence test is that an evaluator should not hold equity in the company or be financially or operationally dependent on it, but the post does not say how METR is funded [21]. In our view the working arrangement is where that test is hardest to check from outside, because the six use Anthropic's hardware and office space [3]. Other METR staff are siloed from the team's work, including researchers who work with Anthropic on other projects [15]. Anthropic describes that as a limit on who can obtain its information [19].
Business Insider reported that the step fulfils a promise Amodei made last month in an essay [13]. "Anthropic is unilaterally committing to this step now," Amodei said at the time [11]. METR did not immediately respond to Business Insider's request for comment, and the six researchers' names were not published [17]. The terms are in place this quarter. How they hold up cannot be seen until METR publishes its analysis [8].
What to watch
- Whether METR's first analysis under the arrangement characterizes a redaction, and whether it calls one unjustified.
- Whether SecureBio's and Accenture's publication rights widen from the more limited starting point Anthropic describes.
- Whether Anthropic or METR discloses how METR is funded, which bears on the equity-and-dependence test.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence68
- Adoption15
- Hype gap+15
- Incentives60
- Confidence72
Perspective Coverage
9 publishers- Builder
- Builder 16%
- Operator
- Operator 49%
- Investor
- Investor 35%
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Anthropic said it has started to give employee-like access to a team from METR, a nonprofit that evaluates frontier AI models.
- [2]
Anthropic signed the White House Accord on Super Intelligence, in which companies pledged to partner with independent external auditors or evaluators.
- [3]
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 have.
- [4]
Outside a small number of clearly defined categories (e.g., data about customers, attorney-client privileged information, or information about model architectures), Anthropic employees are encouraged to share information with the METR team as they would with a fellow employee.
- [5]
Anthropic cannot veto publication of METR's analysis of its risk levels, incidents or practices, but can require METR to redact information that falls into narrow, predefined categories.
- [6]
Anthropic is also working with SecureBio and Accenture as evaluators and is in early discussions with several other organizations, including Hugging Face.
- [7]
Because the other evaluator partnerships are in early stages, those evaluators receive more limited information and have more limited publication rights.
- [8]
METR will publish the results of its risk assessments of Anthropic.
- [9]
Business Insider reported that Anthropic's disclosure is the first time a frontier AI company has taken the step of embedding evaluators.
- [10]
Sam Altman said OpenAI would also embed evaluators after Dario Amodei's essay, though Business Insider reported that has not happened yet.
- [11]
"Anthropic is unilaterally committing to this step now," Amodei said at the time of embedding safety evaluators.
- [12]
Business Insider described the evaluators as having free rein to publish, with a couple of caveats prohibiting the METR team from disclosing customer data, model architecture and attorney-client privileged information.
- [13]
Business Insider reported that embedding evaluators fulfils a promise first made by Amodei last month, when he published an essay calling on frontier AI companies to slow down.
- [14]
Anthropic says bank supervision operates on similar logic, with regulators assigning resident examiners to work inside the largest banks, though it says the analogy is not exact.
- [15]
The engagement is limited to a small METR team under enforceable agreements; other METR employees, including researchers who work with Anthropic on other projects, are siloed from that team's work.
- [16]
"We aim to close the gap between what evaluators understand about a frontier AI company's systems and what the company itself understands."
- [17]
The names of the six METR researchers were not immediately published, and METR did not immediately respond to Business Insider's request for comment.
- [18]
Anthropic said it cannot publicly share all information relevant to its safety work because some of it could be useful to malicious actors or is commercially sensitive, like compute or training information.
- [19]
Anthropic said embedded evaluation is not a way for other organizations to obtain its information, or that of its customers and partners.
- [20]
Over the past year Anthropic has worked with METR on several reviews of its Risk Reports.
- [21]
Anthropic lists independence as a requirement for evaluators: they should not hold equity in the company they evaluate or be financially or operationally dependent on it.
- [22]
Business Insider reported that Anthropic chose the six METR evaluators who receive employee-like access.
Sources
9 independent publishers whose own reporting we read for this story.
- anthropic.comAnthropic says it is giving employee-like access to METR as part of its pledge to partner with evaluators for assessment of its models and safety practices
1 article · October 15, 2026
- blockchain.newsAnthropic expands independent AI evaluator access with METR - Blockchain.News
1 article · October 15, 2026
- businessinsider.comAnthropic just opened its doors to outside safety watchdogs. Here's how it will work.
1 article · October 15, 2026
- channelnewsasia.comAnthropic grants outside evaluator METR employee-level access in AI safety push
2 articles · October 15, 2026
- cnbc.comAnthropic taps nonprofits METR and SecureBio to help assess risks and safety practices
1 article · October 15, 2026
- cryptobriefing.comAnthropic grants METR access to AI models for safety and transparency push
2 articles · October 15, 2026
- cryptopolitan.comAnthropic lets METR publish safety findings without its approval
1 article · October 15, 2026
- techcrunch.comAnthropic is letting METR researchers look under the hood with employee-level access
1 article · October 15, 2026
- zdnet.comAnthropic gives third-party auditors ’employee-like access’ in AI industry first
1 article · October 16, 2026
Topics and entities
Topics
- US AI regulation
- Third-party AI audits
- AI Safety Evaluation