Why METR’s ties are raising questions about Anthropic’s AI watchdog plan

Questions about METR’s independence intensified on September 15–16, 2026, after Dario Amodei’s proposal for embedded third-party AI evaluators was linked to the nonprofit’s donor, staff and institutional networks. METR says it does not accept funding from AI companies or their employees and that its funders have no say over its projects. The dispute is about whether those safeguards are persuasive enough for a watchdog role—not about a demonstrated failure in a specific evaluation.

Why METR’s independence became an issue

Amodei, Anthropic’s CEO, proposed that frontier AI companies give outside evaluators ongoing access comparable to that of employees. The evaluators would examine safety practices, incident reporting, model alignment, training pipelines and internal processes rather than relying only on information selected by the companies.

METR was the organization named in discussion around that proposal. The nonprofit’s suitability is now being questioned because of reported overlaps involving its predecessor, staff members, Effective Altruism networks and philanthropists associated with Anthropic investors.

That overlap does not amount to a finding that Anthropic directs METR. It does, however, create a straightforward governance question: can an evaluator be seen as independent when people and funding networks connect it to the company whose work may come under scrutiny?

What Dario Amodei proposed

Amodei’s proposal, published on September 12, called for embedded third-party evaluators with “employee-like access” inside leading AI companies. In practice, that would mean access to the tools, permissions, training pipelines and safety work needed to investigate how advanced models are developed and operated.

The proposed scope includes:

• reviewing model alignment and safety commitments;
• examining incident reports and internal processes;
• assessing training pipelines; and
• publishing important findings, subject to narrow redactions for security, legal, commercial or third-party confidential information.

The model is designed to give external investigators a view closer to the one available to internal risk teams. Amodei said Anthropic was committing to the approach and called for other frontier AI companies to adopt it. Sam Altman separately supported independent evaluators with employee-like access, but his endorsement did not specifically name METR.

The documented network connections

METR grew out of the Alignment Research Center, or ARC. The organizations are associated with Paul Christiano, who worked with Amodei at OpenAI in the 2010s and later served as one of the first five trustees of Anthropic’s Long-Term Benefit Trust.

The reported relationships also involve funding networks connected to Anthropic investors Dustin Moskovitz and Jaan Tallinn. The Survival and Flourishing Fund, associated with Tallinn, reportedly earmarked up to $752,000 for METR since its founding: $324,000 in general cash and a $428,000 matching pledge.

Coefficient Giving, associated with Moskovitz, reportedly gave $1,515,000 to ARC in 2022. METR says that donation was firewalled and was not used for its operations.

Separately, METR announced approximately $71 million in funding commitments raised during the six months before its August 14 announcement. METR says it does not accept funding from AI companies or their employees and takes no compensation from AI labs.

The important distinction is between direct corporate funding and overlapping philanthropic or personal networks. The reported relationships concern the latter. METR’s stated policy concerns the former, as well as control over its projects.

A short timeline of the dispute

What METR says about its role

METR says its funders have no say in the projects it works on. It also says it does not accept funding from AI companies or their employees. In response to the independence criticism, the organization presented external investigators as a way to bring more information from inside AI companies into the public domain.

Critics David Sacks and Perry Metzger argue that the overlap among donors, staff and institutions makes claims of independence misleading. Their comments are criticisms of METR’s structure and relationships, not findings by a regulator or court.

The disagreement therefore turns on more than a grant ledger. An evaluator’s independence depends on who can influence its projects, how conflicts are disclosed, whether access is protected from company control and whether important findings can be published. METR’s public position addresses funding and project control; the critics are challenging whether those safeguards are sufficient in a network where professional and philanthropic relationships overlap.

What the dispute establishes

The most concrete facts are these: Amodei proposed embedded evaluators with employee-like access; METR is the organization associated with that oversight discussion; METR has reported funding commitments of about $71 million; and documented relationships connect METR’s wider institutional and philanthropic network with people associated with Anthropic.

Those facts explain why the watchdog proposal has drawn scrutiny. They do not show that Anthropic directly funded METR, that Anthropic controlled its projects, or that a particular METR evaluation was compromised. METR’s position is that AI-company funding is excluded and that donors cannot direct its work.

The practical test for the proposal would be whether evaluators receive meaningful access, can investigate incidents independently and can publish important findings without the company deciding what the public gets to see.