OpenAI vs NYU Professor: Did AI Really Solve the Navier-Stokes Problem, or Steal It?

OpenAI's $1M Math Breakthrough Is Now a Credit-Stealing Scandal

What should have been one of the biggest wins in AI history has turned into an ugly fight over stolen credit.

On Tuesday, OpenAI announced that its AI had solved the Navier-Stokes existence and smoothness problem — a math puzzle that's stumped researchers for nearly a hundred years. It's one of seven "Millennium Prize Problems," each carrying a $1 million bounty from the Clay Mathematics Institute. According to OpenAI, an unreleased model more powerful than GPT-6 Astra ran 10,000 AI agents for 88 straight hours to crack it, at a compute cost running into the millions of dollars.

That would be huge news on its own. But hours before OpenAI's announcement, NYU mathematician Tristan Buckmaster published a four-page statement telling a very different story.

What Buckmaster Says Actually Happened

Buckmaster had spent nearly a year working on a related version of the same problem with Levent Alpöge, a mathematician who works at Anthropic — OpenAI's competitor. This wasn't a company project. It was personal, self-funded work, and Buckmaster says he even paid out of pocket for a hefty OpenAI bill while storing his drafts in OpenAI's Codex coding tool.

By late August, the pair believed they'd found a solution. Buckmaster reached out to an OpenAI mathematician after hearing rumors that a rival lab had made a breakthrough — just to check what was going on.

That's when things got strange. Just days later, OpenAI told him they'd independently solved the full Navier-Stokes problem using the exact same rare approach Buckmaster says almost nobody else in the field was even attempting. "It is not the direction one arrives at in a few days," he wrote.

When Buckmaster pushed to ask whether OpenAI's model had been trained on his Codex sessions — where all his unpublished work was stored — he says the question went unanswered.

The Career Threat

According to Buckmaster, OpenAI's Sébastien Bubeck, who leads the company's math team, offered him a deal: publish jointly with OpenAI, but drop Alpöge from the credit since he works for a competitor. When Buckmaster said he'd rather publish independently and go public with what happened, he says Bubeck told him:

"Why would you ruin your career?"

And when Buckmaster pushed back further, Bubeck allegedly added: "If you don't want me to be nice, then I don't have to be nice."

OpenAI's Response

OpenAI, Bubeck, and CEO Sam Altman all deny the accusation. Bubeck posted screenshots of texts he says show he offered a fair, coordinated release and never asked for Alpöge to be removed from his own paper — only that Buckmaster might write up a separate account of OpenAI's version. Bubeck admitted his "ruin your career" comment was a poor choice of words, but said he only meant to warn Buckmaster about making unfounded claims publicly.

Altman backed Bubeck fully, saying he "acted with integrity and generosity throughout."

As for the data question, OpenAI's official statement is carefully worded: the company says it did not access any of Buckmaster's specific user data, but admitted it "cannot rule out" that anonymized data from his usage may have helped improve its models.

Why This Actually Matters

Buckmaster has been careful not to accuse anyone outright. "I do not know what their model did, or how. I do not know whether our data was used. I am not accusing anyone of anything," he wrote.

But the incident raises questions far bigger than one math proof:

• Do AI companies train on your data, even when you think it's private or personal?
• Who gets credit when an AI model produces work that closely mirrors a paying customer's own unpublished research?
• Are AI labs racing to solve problems because of genuine science, or because of fear of losing a headline to a competitor?

Even mathematician Terrence Tao weighed in this week, warning that AI models "harvesting" open problems for quick answers could damage the very ecosystem that trains the next generation of mathematicians.

The Clay Mathematics Institute still officially lists the Navier-Stokes problem as unsolved and says it will conduct a detailed review before recognizing any solution. OpenAI has said it won't claim the $1 million prize regardless of outcome.

For now, the science is unresolved — and so is the trust between two of the world's biggest AI companies.