• OpenAI's new nine-mathematician Advisory Group on Mathematics and AI, hosted at the Institute for Advanced Study, will review how AI-generated math claims are assessed and disclosed — but its charter explicitly excludes advising on release pacing.
• The panel follows a credit dispute over OpenAI's September 8, 2026 Navier-Stokes proof: NYU's Tristan Buckmaster's own public statement alleges OpenAI researcher Sébastien Bubeck pressured him to drop Anthropic mathematician Levent Alpöge as co-author; Bubeck has publicly denied the allegations on X as false and inflammatory.
• OpenAI simultaneously announced its internal model has resolved 100+ additional open problems, a claim that shipped before the new panel had reviewed anything.
• 25 Fields Medal-winning mathematicians signed an open letter warning that AI labs are treating unsolved problems as competitive benchmarks at the expense of academic norms, per TechCrunch.
• Panel members serve unpaid and can publicly criticize OpenAI, giving the group real but narrow independence — undercut by having no say over announcement timing.

OpenAI on September 21, 2026 announced the Advisory Group on Mathematics and Artificial Intelligence, a nine-mathematician panel hosted at Princeton's Institute for Advanced Study that will review how the company assesses and publicly communicates AI-generated math results, according to OpenAI's own announcement and TechCrunch's reporting on it. The panel is unpaid, independent, and free to criticize OpenAI in public — but it arrives explicitly barred from touching the one decision that mattered most in the two weeks before it existed: how fast OpenAI announces a result.

A credit fight with two named sides

The panel follows a dispute triggered by OpenAI's September 8, 2026 announcement that an internal system had proved the Navier-Stokes existence-and-smoothness problem, a Clay Mathematics Institute Millennium Prize problem. NYU mathematician Tristan Buckmaster alleges OpenAI researcher Sébastien Bubeck pressured him to drop his collaborator, Anthropic mathematician Levent Alpöge, from co-author credit — telling him, in Buckmaster's own public statement, 'Why would you ruin your career?' Bubeck has publicly denied the allegations on X as 'false and inflammatory,' saying he 'came into the discussion following academic norms,' per Fortune's and TechCrunch's reporting on the exchange.

What the panel can actually review

OpenAI's charter, as reported by TechCrunch, tasks the nine members with advising on how the company reviews emerging math results, assesses their significance, coordinates how they're disclosed, and applies academic publishing norms before a claim goes public. Confirmed founding member Camillo De Lellis of the Institute for Advanced Study joins a roster OpenAI describes as senior mathematicians drawn from outside the company. Members are unpaid, can offer unsolicited advice, and are free to publish their own criticism of OpenAI — giving the group real, if narrow, independence from the lab it advises.

The one power it doesn't have

OpenAI states directly, per TechCrunch's report on the announcement, that 'the group will not be responsible for advising us on how to pace our internal progress on mathematics' — and the Institute for Advanced Study confirms the panel 'will not have decision making power.' That is exactly the mechanism at the center of the Buckmaster dispute: OpenAI, not outside reviewers, still decides the day a claim ships, peer review or not. A board with no say over launch timing does not stop a repeat of an unreviewed announcement; it gives critics a more credentialed venue to object afterward.

A bigger, unreviewed claim shipped the same day

Alongside the panel, OpenAI said its internal model has now resolved more than 100 additional open problems across most areas of mathematics, TechCrunch reported. The announcement came days after 25 Fields Medal-winning mathematicians signed an open letter warning that AI labs racing to claim high-profile mathematical results threaten the integrity of academic mathematics. None of those individual results has been detailed for outside review, and the newly named advisory panel has not yet reviewed any of them.

The bottom line

This is a real governance fix aimed at the wrong failure. OpenAI's math credibility problem was never a shortage of qualified reviewers — Buckmaster and the wider math community scrutinized the Navier-Stokes claim for free, in public, within days of the announcement. The problem was that OpenAI shipped the claim, and reportedly discussed cutting a collaborator from credit, before that review happened. A panel explicitly barred from advising on publication pace does not close that gap. The number to watch: how many of OpenAI's next math claims clear this group's review before the press release, not after — as of launch, that number is zero.