Mathematician Tristan Buckmaster believes OpenAI used his work to jump ahead and beat him to solving the legendary Navier–Stokes existence and smoothness problem, which carries a $1 million prize.

But that hasn’t stopped Buckmaster from using OpenAI’s model. He’s not the only mathematician who feels conflicted about relying on artificial intelligence.

“Even if you don’t agree with any of it, people get stuck. AI is so useful that it’s hard to completely prevent its use,” Buckmaster told WIRED. “These companies have monopolies and don’t have much choice,” he added.

Why Buckmaster Believes OpenAI Used His Research

In the week and a half since the New York University professor accused OpenAI of copying his approach, Buckmaster has been using the company’s coding agent, Codex, to organize his research papers. When he has time to do mathematics—which he says is rare given the attention surrounding the dispute—the tool helps him understand the logical steps OpenAI’s agents may have taken from his initial work to its final proof.

Buckmaster and Anthropic researcher Levent Alpöge were working on the Navier–Stokes problem using Codex and Anthropic’s competing models. Buckmaster said OpenAI deployed tens of thousands of agents to arrive at a solution, but only after learning that the equation was close to being solved.

Buckmaster’s publication of his claims sparked a firestorm over artificial intelligence and whether AI could make human mathematicians obsolete. The controversy also prompted OpenAI to investigate and address the issue. In its announcement, OpenAI said that “the two months of Buckmaster’s Codex prompts preceding this announcement and the September 8, 2026 paper did not impact the system in any way, including training.” The company pointed WIRED to the announcement in an email.

AI’s Growing Role in Mathematical Research

Showing that AI is pushing the boundaries of mathematics was “more important than the results,” Buckmaster said. But he believes it is irresponsible and “childish” to produce solutions to age-old mathematical problems without fully acknowledging the human work behind them, particularly as the company prepares for a major IPO.

Other mathematicians have raised similar concerns. Only a handful of people around the world understand the techniques in geometric group theory that German mathematician Andreas Thom has developed over the past 20 years. So when OpenAI said its Astra model had used those techniques to prove a long-standing problem Thom had been working on, he was surprised.

“And of course I wondered: How did they know that?” Thom said.

Thom asked OpenAI researchers Mark Sellke and Sébastien Bubeck about the result. In an August email, he noted that the company’s claim that “no progress” had been made on the problem over the previous decade overlooked a 2019 paper by Thom and other mathematicians. OpenAI later revised its press release.

Thom said that, in the months leading up to the results, he and one of his colleagues had used ChatGPT while working on the problem. When he asked whether those interactions had been included in the model’s training data, Sellke replied, “Not at all.”

“I put that aside,” Thom recalled. “I’m not really interested in politics. I want to work on mathematics.”

Can AI Contributions to Mathematics Be Traced?

Thom has seen OpenAI’s statement that Buckmaster’s prompts could not have affected its system, but he says he does not believe it and admits he may never know whether his work influenced the results.

“AI completely obliterates the idea that you can track who contributed what,” he said. “Maybe it’s over.”