Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: OpenAI says a model of its resolved more than 100 open math problems this year, including an abrupt claimed solution to a Navier-Stokes Millennium Prize question in early September. The company has now formed a nine-person mathematical advisory group, but 25 Fields Medal winners already signed a letter warning that AI labs are racing to claim famous problems faster than the math community can verify them.
Solving one of the seven Millennium Prize Problems, the set of math questions so hard the Clay Mathematics Institute put a million-dollar bounty on each of them, would normally be the biggest story in mathematics for a decade. OpenAI’s claimed progress on Navier-Stokes, the equations that describe how fluids flow, showed up abruptly in early September with little warning, as one item in a list of more than 100 problems the company says its model has resolved this year.
That pace is either a genuine leap forward for mathematical research or a sign that verification hasn’t kept up with output, and right now a lot of professional mathematicians suspect it’s some of both.
Nine advisors, and only one who signed the warning letter
OpenAI’s response has been to stand up a nine-person mathematical advisory group made up of academics from outside the company. Only one of the nine, Camillo De Lellis of the Institute for Advanced Study, also put his name on a separate open letter signed by 25 Fields Medal winners, the field’s equivalent of a Nobel Prize, warning that AI labs are threatening the norms of mathematical research by racing each other to claim famous open problems.
“This group will serve as a bridge to the mathematical community and broader public, giving mathematicians a voice in how we move forward,” OpenAI said in announcing the advisory panel.
The Institute for Advanced Study, one of the most prestigious math research centers in the world, was notably more careful about what the arrangement actually means. In its own statement, the institute said plainly: “Although we will give advice, we do not have decision making power at any AI company.” That’s a significant caveat buried in polite language. The advisory group can weigh in, but it has no ability to slow down or redirect OpenAI’s ongoing mathematical research, a limitation OpenAI itself has acknowledged.
Why 25 of the world’s best mathematicians are nervous
The concern isn’t really about whether AI can do good math. Plenty of mathematicians already use AI tools to check proofs or explore conjectures. The concern is about incentives. If AI labs treat famous unsolved problems as PR opportunities, racing to be first and loudest rather than most rigorous, the traditional peer review process that catches errors in complex proofs gets skipped or rushed. A wrong proof of a hard problem can take years to properly disprove, especially if it runs hundreds of pages and nobody wants to be the one who spent six months finding the flaw.
There’s also a quieter worry about credit and culture. Mathematics has always been a slow, collaborative, heavily scrutinized field where reputations are built over careers, not news cycles. An AI lab claiming a Millennium Prize breakthrough in a press release, before independent mathematicians have had time to fully verify the work, cuts against everything that field is built on.
Bottom Line: An advisory group with no decision-making power isn’t oversight, it’s a communications buffer. If OpenAI’s Navier-Stokes claim holds up under real scrutiny, this will be remembered as a legitimate breakthrough. If it doesn’t, the advisory panel gives OpenAI a group of respected names to point to anyway, which is probably the actual point of creating it right now.
