OpenAI Claims AI Solution to Navier-Stokes Millennium Prize Problem

OpenAI has announced that an internal AI system discovered a solution to the Navier-Stokes existence and smoothness problem. The breakthrough addresses a 90-year-old mathematical challenge regarding the behavior of fluid dynamics.

The company reports that its AI produced both an analytical proof and a Lean formalization demonstrating that 3D incompressible fluid motion can develop a singularity in finite time. To achieve this, OpenAI deployed a multi-agent system consisting of approximately 10,000 concurrent agents, which generated roughly 130 billion output tokens specifically for the Navier-Stokes problem. The solution was reached on September 5, with formal verification via Lean completed by September 6. According to OpenAI's Mark Chen, the computing costs for the discovery reached into the millions of dollars.

The Mathematical Stakes

The Navier-Stokes equations are fundamental to understanding how water and air move. Since 2000, the problem has been one of seven Millennium Prize Problems established by the Clay Mathematics Institute, each carrying a $1 million reward. The central question is whether smooth fluid motion can "break down" into a singularity—a point of infinite speed—within a finite amount of time. Solving this would provide a definitive answer to one of the most enduring questions in physics and mathematics.

A Dispute Over Credit

Despite the technical achievement, the announcement has been met with allegations from mathematician Tristan Buckmaster of NYU and researcher Levent Alpöge of Anthropic. The pair released documents claiming they had made key advances in related "forced Euler" equations around the same time as OpenAI's discovery. Buckmaster alleges that OpenAI offered several proposals to handle the discovery, including one where he could publish a paper announcing the solution as an internal OpenAI discovery while omitting Alpöge's name.

OpenAI has denied accessing private user data to solve the problem, though the company noted that de-identified data from product usage may have helped improve the underlying models. OpenAI maintains that its solution was independently derived and mathematically distinct from the work of the other researchers.

Implications for AI Research

This event underscores the accelerating role of artificial intelligence in frontier mathematics and the potential for "credit wars" as AI speeds up the pace of discovery. It also raises critical questions regarding data privacy and whether AI labs might inadvertently or intentionally use user-generated prompts to steer their own research breakthroughs. As AI systems move from assisting researchers to independently generating proofs for Millennium Prize problems, the industry must now grapple with how to attribute intellectual property in an era of automated discovery.