NYU Mathematician Accuses OpenAI of Scooping Millennium Prize Proof

NYU mathematics professor Tristan Buckmaster and Anthropic mathematician Levent Alpöge claim OpenAI used their unpublished research to solve the Navier-Stokes existence and smoothness problem. The dispute centers on whether the AI lab leveraged user data and brute-force computing to scoop a career-defining mathematical breakthrough.

The Navier-Stokes problem is one of seven Millennium Prize challenges, carrying a $1 million bounty. Buckmaster and Alpöge had been quietly attacking the problem using a specific, uncommon mathematical tactic involving "smooth force." According to Buckmaster, almost no one else he knew of was working on this specific route. OpenAI subsequently adopted the same tactic to reach a formal proof. OpenAI claims the discovery was made by an unreleased next-generation model that utilized 300 billion output tokens, a process the company estimates cost approximately $22.5 million in compute resources.

The Collision of AI and Academia

The Navier-Stokes equations are fundamental to fluid mechanics, describing how liquids and gases move. Despite their importance, they have remained theoretically unsolved for decades. This case highlights a growing tension between traditional academic research and the immense resources of AI labs. While mathematicians typically rely on theoretical insight and peer review, AI labs can now apply "compute-brute-forcing" to explore mathematical paths at a scale impossible for human researchers.

Intellectual Property Risks

The controversy raises critical questions about the vulnerability of researchers who use AI tools for their work. Buckmaster expressed concerns that OpenAI's models may have absorbed his proprietary work through his own usage of the tool. In an official response, OpenAI denied accessing specific user data to solve the Navier-Stokes problem. However, the company admitted it cannot rule out that de-identified data from general product usage helped improve the models that eventually found the proof.

Industry Implications

This dispute underscores a profound shift in intellectual property risks in the age of Large Language Models (LLMs). If proprietary, unpublished breakthroughs can be absorbed by a model provider and then replicated by the provider's own systems, the incentive for researchers to use these tools diminishes. The case suggests a blurred line between algorithmic "inspiration" and the systemic absorption of user-contributed intellectual property.

What Remains Unconfirmed

While the technical overlap in the "smooth force" tactic is evident, the exact mechanism by which OpenAI's model arrived at the proof remains a subject of debate. It is not yet clear if the model independently converged on the tactic or if the training data—potentially including user inputs—steered it toward Buckmaster's specific approach. The academic community continues to watch whether this will lead to new standards for how AI labs handle sensitive research data.