OpenAI claims a Navier-Stokes proof, sparking a math dispute

OpenAI announced on September 8 that one of its internal models had produced a proof addressing one of the Millennium Prize Problems: the existence and smoothness of solutions to the Navier-Stokes equations. These equations describe fluid motion — from the vortex in a bathtub drain to airflow around an aircraft wing to blood flow in vessels — and have been studied for more than two centuries without anyone knowing whether their solutions always stay smooth, regardless of starting conditions. OpenAI says its model proved that, under certain conditions, the equation's solution diverges to infinity in finite time — a phenomenon known as "blow-up". The problem is one of the Millennium Prize Problems carrying a $1 million reward from the Clay Mathematics Institute, but the version OpenAI proved includes a forcing term that excludes it from the prize's criteria; the company said it would not claim the award.

10,000 AI agents in 88 hours

According to OpenAI, the proof was produced by its internal model using roughly 10,000 AI agents working in parallel over about 88 hours, then verified in the Lean formal-proof language. Just before that announcement, New York University mathematician Tristan Buckmaster had posted on social media, together with mathematician Levent Alpöge — an Anthropic employee — a proof of the same kind of blow-up for the Euler equations, a simpler relative of Navier-Stokes. In a statement released afterward, Buckmaster alleged that OpenAI, having heard rumors of the pair's progress, adopted the same method to extend the proof to the full Navier-Stokes equation. He said that during a call with OpenAI researcher Sébastien Bubeck, he was offered two options: remove Alpöge from the author list because of his Anthropic affiliation, or let OpenAI publish its result alone.

OpenAI's denial and the mathematicians' unease

OpenAI researcher Sébastien Bubeck firmly denied the allegations. He said the company's internal model never had access to Buckmaster and Alpöge's work, and that OpenAI's approach to the Euler equations differs fundamentally from theirs. He did acknowledge, however, that the final proof for Navier-Stokes followed a path similar to the method the two mathematicians used. In its official statement, OpenAI said it never viewed their work before it was made public, and that no specific user data was used to produce the proof — while not ruling out that de-identified data from the pair's use of its products may have indirectly helped improve its models. OpenAI CEO Sam Altman also weighed in on social media, saying he would have preferred better coordination.

• Announcement date: September 8, 2026
• Method: roughly 10,000 AI agents, about 88 hours of computation
• Verification tool: the Lean formal-proof language
• Rival team: Buckmaster (NYU) and Alpöge (Anthropic) first proved blow-up for the Euler equations
• Millennium Prize: $1 million from the Clay Mathematics Institute, not applicable to this version of the problem

The episode goes beyond a simple priority dispute. Mathematician Tristan Buckmaster compared it to the moment IBM's Deep Blue supercomputer beat chess champion Garry Kasparov in 1997, saying the scientific community now needs a careful, serious discussion about AI's role in research. University of Sydney mathematician Zsuzsanna Dancso said OpenAI's conduct ran against the basic ethical principles of mathematics: acknowledging sources and collaborators. Crucially, the proof has not yet been through independent peer review, leaving open a basic question — is this a genuinely verified result, or a claim built for announcement effect?

For the wider mathematics community, the episode is a reminder that speed and spectacle are not the same as verification. Universities and funding bodies that are beginning to back AI-assisted formal proof work will need clear rules on authorship and data provenance before the next such race breaks out between labs — rules that, as this dispute shows, none of the major AI companies has settled yet. Until the Navier-Stokes proof clears independent peer review and formal checking, it remains a strong claim rather than a settled result.