OpenAI announced Tuesday that an unreleased model solved the Navier-Stokes problem, one of mathematics' seven Millennium Prize questions, in 88 hours using a swarm of roughly 10,000 AI agents. The problem, which concerns the movement of fluids, has stumped human researchers for close to 90 years and carries a $1 million bounty from the Clay Mathematics Institute. The result is a genuine technical achievement. It is also mired in allegations of scooping, data access concerns, and conduct that mathematicians say violates the informal norms their field depends on.
A Rushed Effort Triggered by Twitter Rumors
By OpenAI's own account, the effort was a hurried and expensive affair, costing millions of dollars. The company said it began working on the problem after hearing rumors on Twitter that other researchers were making progress on Millennium Prize problems. Sébastien Bubeck, an OpenAI researcher, said at a press briefing that the company thought: "We have such a strong model. Why don't we try to solve also a Millennium Prize problem?"
OpenAI said it only later realized the rumors concerned Levent Alpöge, a researcher at Anthropic, and Tristan Buckmaster, a mathematics professor at New York University. Alpöge was not working on behalf of Anthropic. The company has said it does not intend to claim the bounty, which has not yet been awarded, and that its only goal is to report on the progress of its AI models.
But the timeline raises questions. OpenAI does not appear to have publicly discussed Navier-Stokes work before September. The company threw significant resources at the problem after hearing about other researchers' progress, then announced its result one day before Buckmaster and Alpöge published their own findings on a related problem.
Allegations of Data Access and Intimidation
Buckmaster contacted OpenAI after learning the company had become aware of his and Alpöge's progress. He asked when OpenAI began working on the problem and what data its model had been trained on. According to Buckmaster, the conversation turned sour quickly. An OpenAI researcher asked him, "Why would you ruin your career?" when he said he would go public. When Buckmaster asked why going public would ruin his career, he received the reply: "If you don't want me to be nice, then I don't have to be nice."
OpenAI urged Buckmaster to publish the work and credit OpenAI's internal model while dropping Alpöge as coauthor. Buckmaster asked whether OpenAI had accessed his sessions on Codex, which he had used while working on the problem. The company grew increasingly evasive, he said.
OpenAI has flatly denied using any specific user data. In its blog post, the company said: "We (the researchers and the agents) did not see any of their work through any means until they released it publicly." But OpenAI could not conclusively rule out indirect influence. "While unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models," the company said, while stressing that the two proofs differ significantly.
Mathematicians Call the Conduct a Violation of Trust
The math community's reaction has been sharp. Abhishek Saha, a mathematics professor at Queen Mary University of London, described OpenAI's actions as "the kind of things that mathematicians will generally not do." Matthew Ballard, a professor at the University of South Carolina and associate director at the Institute for Computer-Aided Reasoning in Mathematics, said: "Mathematics depends heavily on an informal norm of trust. Researchers routinely share incomplete ideas and ongoing work with colleagues to sharpen their thoughts. It is done with the expectation that it will not turn into a competition."
Jeremy Avigad, a Carnegie Mellon professor and ICARM director, said that even "the thought that AI systems might steal ideas from our queries is chilling." He added: "Now that even the slightest hint might be enough for someone with sufficient computational resources to set a swarm of agents on solving the problem, people are likely to be more cautious. It's sad to think about how that might change the research environment."
Brendan Hassett, a Brown University professor, said OpenAI's inability to say whether its models were informed by other mathematicians' work compounds the unease. "Given the history of the AI companies appropriating copyrighted work without permission or payment, it is natural for people to ask these questions," he said. Companies should be held accountable to demonstrate that chat logs will not be used to improve models.
A PR Victory That May Cost More Than It Gains
Andras Juhasz, an Oxford professor, called the result "clearly a PR victory for OpenAI" but questioned its sustainability. He noted that human mathematicians scoop each other too, though what OpenAI did shows this can happen on a much grander scale. "Suddenly, 10,000 mathematicians jump on your problem," he said.
Yang-Hei He, a fellow at the London Institute for Mathematical Sciences, said he worries that "maths under the big companies is much too secretive." As someone who describes himself as "always optimistic about AI," he admitted concern that mathematics could be reverting to a more secretive state, similar to the era when it was funded by patronage from wealthy families like the Medicis.
The broader question is what this means for the relationship between AI companies and academic research. OpenAI has proven its models can compete at the frontier of mathematics. In doing so, it appears to have alienated the community it has been trying to impress. The trust that underpins collaborative research, where mathematicians share incomplete work without fear of being scooped, may be difficult to restore once broken. And the precedent of deploying massive compute against problems where other researchers are known to be working sets a dynamic that the field has not had to contend with before.