OpenAI announced that one of its AI systems has solved the Navier-Stokes problem, one of mathematics' most famous unsolved puzzles worth $1 million. The achievement is remarkable, but it has sparked an unexpected controversy about whose work actually solved it, how AI training data influences breakthroughs, and what happens to the traditional pace of scientific discovery when thousands of AI agents can work simultaneously on the same problem.

What Is the Navier-Stokes Problem, and Why Does It Matter?

The Navier-Stokes equations describe how fluids move, whether that's water flowing through a pipe, air moving around an aircraft, or smoke swirling through a room. For roughly 90 years, mathematicians have tried to prove whether these equations always behave predictably or whether they can suddenly break down and produce what's called a singularity. In 2000, the Clay Mathematics Institute put it on a list of seven exceptionally difficult problems and attached a $1 million prize to each one.

The problem matters because understanding fluid dynamics at this fundamental level has applications across engineering, physics, and climate science. Yet despite decades of effort, no one had cracked it, until now.

How Did OpenAI's AI Solve It?

OpenAI deployed approximately 10,000 AI agents that worked on different parts of the problem simultaneously. Over roughly 88 hours in early September, these agents collectively generated about 2.7 million messages and 130 billion output tokens. The company estimates the equivalent computing power would have cost around $15 million at standard customer prices.

The AI found a mathematical construction in which a smooth fluid could develop a singularity in finite time. OpenAI then had another model spend about 17 hours translating the result into Lean, a formal language that lets computers verify mathematical proofs automatically. The company published a 165-page write-up alongside the formal verification.

Why Is There a Credit Dispute?

Here's where the story gets complicated. Two mathematicians were working on closely related ideas at almost exactly the same time. Tristan Buckmaster, a mathematics professor at New York University, and Levent Alpöge, a mathematician at Anthropic (OpenAI's rival), had spent months on the problem privately. Buckmaster had been using OpenAI's Codex, a coding tool, to help with his research, uploading his work and prompts directly into the system.

When OpenAI learned about their progress in early September, the company accelerated its own effort. Buckmaster then asked OpenAI a critical question: could his unpublished research, stored in his Codex sessions, have somehow influenced the company's AI model and contributed to the solution?

OpenAI's response was cautious. The company said its internal system did not have access to his private Codex sessions and later claimed an investigation found that his prompts in the two months before the announcement could not have influenced the system, including through training. However, OpenAI acknowledged it could not rule out the possibility that user data contributed to earlier model training.

What Does This Mean for How Science Works?

The Navier-Stokes dispute reveals a fundamental shift in how mathematical progress happens. For centuries, mathematical discovery has followed a familiar rhythm: someone has an idea, works on it for years, discusses it with colleagues, writes proofs, and other mathematicians critique the work. Eventually, if everyone agrees it holds up, the result becomes part of human knowledge.

Now imagine replacing some of those steps with 10,000 AI agents working simultaneously for 88 hours. The speed is unprecedented, but so is the confusion about credit and attribution. Consider the key questions this raises:

• Data Transparency: Most users never check the "Improve the model for everyone" toggle when uploading work to AI tools, meaning their research may be used to train future models without explicit consent.
• Incentive Structures: If a rumor about progress on a hard problem can trigger massive AI effort that "flattens" the original research before it reaches completion, researchers may stop sharing ideas early, a practice that has been central to mathematical progress for centuries.
• Attribution and Credit: When an AI discovers the proof, who gets the credit: the company that built the model, the researchers who guided the AI, the mathematicians whose earlier work appeared in training data, or the machine itself?

"A rumor alone can now trigger enough AI effort to flatten it before the original research project has time to reach its full potential," noted mathematician Terence Tao on Mathstodon, highlighting how the incentive structure for sharing ideas in mathematics has fundamentally changed.

Terence Tao, Mathematician at UCLA

How to Protect Your Research When Using AI Tools

The Navier-Stokes case offers practical lessons for researchers and professionals who use AI systems:

• Check Your Privacy Settings: Review whether your AI tool is set to "improve the model for everyone" or similar data-sharing toggles. Disable them if you're working with sensitive or unpublished research.
• Understand Data Retention Policies: Ask your AI provider explicitly whether your prompts and uploaded work are retained, used for training, or shared with other systems. Get answers in writing.
• Consider Timing and Disclosure: Be cautious about uploading unpublished work to widely used AI tools, especially if you're working on a competitive or high-stakes problem. The rumor effect means others may learn of your progress and accelerate their own efforts.

OpenAI says it won't claim the $1 million prize for solving Navier-Stokes, which is a gracious move. But the bigger question the company has accidentally created is one that mathematics has never really had to confront before: when a machine finds the answer, who exactly solved the problem?

As AI systems become more capable and more integrated into research workflows, this question will only become more urgent. The Navier-Stokes breakthrough is a remarkable achievement, but it may be remembered less for the mathematics and more for the precedent it sets about how science itself is changing.