METR's Ties to Labs Spark Audit Independence Concerns

Episode 1 · METR Investigator's Marriage to OpenAI Board Member Sparks AI Audit Independence Debate (2026-09-12, 6 posts)

AI safety researcher Deborah Raji publicly questioned the independence of AI audits on September 12. While crediting Anthropic for opening up to external audits, she argued that a mature audit ecosystem requires behavioral standards for auditors. She was surprised to find that a METR investigator involved in the HF/OpenAI incident investigation is married to an OpenAI board member, and that most access contract terms also carry restrictions—raising the question of how the current "embedded evaluator" model can guarantee true independence and substantive judgment.

Confirmed

• Raji explicitly pointed out the marriage between the METR investigator in question and an OpenAI board member, and noted that most access contract terms constrain evaluations (details not fully elaborated in the post).
• Replying to Nathan Lambert, Melanie Mitchell raised a broader concern: how can "independent" AI evaluators avoid a revolving door with AI companies— independence is hard to maintain if evaluators hope to be hired by OpenAI or have their startups acquired.
• Nathan Lambert responded that he doesn't know the institutional details and the problem is genuinely hard, but he believes people like himself who have "over-promised independence" for reasons such as blogging are at least a starting point toward independence.

Perspectives

• An industry insider (relayed by emax) complained that so-called independent AI evaluation outfits mostly come from the same EA (effective altruism) circles, funded by the same Open Philanthropy money, with founders who previously worked at OpenAI—hardly truly independent.
• Another poster (relayed by DrAtoosa) took a more moderate stance: METR is currently the closest thing to an independent AI standards body, so its involvement in evaluations is only natural, though they acknowledged that METR members share an ideological echo chamber with OpenAI/Anthropic leadership on safety issues like x-risk—oversight shouldn't all come from one circle.

Why it matters

• External audits are seen as a key mechanism for AI companies' safety governance, and this controversy reveals structural independence problems in the audit ecosystem—personal ties, funding sources, and ideological homogeneity—that could undermine the credibility of audit conclusions. The incident has also fueled discussion of auditor conduct standards and institutional safeguards for independence.

• Deborah Raji flags audit conflict: METR investigator married to OpenAI board member — rajiinio · 2026-09-12
• METR Investigator Married to OpenAI Board Member Sparks AI Audit Independence Concerns — rajiinio · 2026-09-12
• METR as AI Evaluator Makes Sense, But Oversight Needs Ideological Diversity — Dr_Atoosa · 2026-09-13
• Critics say independent AI evaluators share the same EA funding and OpenAI ties — emax · 2026-09-13
• Melanie Mitchell asks how "independent" AI evaluators can escape the revolving door — MelMitchell1 · 2026-09-13
• Nathan Lambert on evaluator independence: overcommitted bloggers are a starting point — natolambert · 2026-09-13

Episode 2 · Critics Allege 'Revolving Door' Between AI Labs and Safety Evaluators (2026-09-13, 2 posts)

Anthropic nominated METR as a third-party evaluator right after an employee jumped to METR, while critics argue safety evaluators like METR and Redwood are staffed by ex-lab insiders holding stock options, creating a 'revolving door' that undermines their independence.

• RedWood/METR evals staffed with ex-lab insiders, warns of a "revolving door" regulatory cartel — beffjezos · 2026-09-13
• Anthropic employee joins METR; next day METR named third-party evaluator — NielsRogge · 2026-09-13