OpenAI Navier-Stokes Proof Sparks Academic Credit Dispute

OpenAI announced on Tuesday that an unreleased AI model has produced a verified proof for the Navier-Stokes existence and smoothness problem, a feat involving 10,000 coordinating agents working over 88 hours [2, 3]. The achievement, which addresses one of seven $1 million Millennium Prize problems, has triggered a public confrontation with NYU mathematician Tristan Buckmaster, who alleges OpenAI leveraged his unpublished research to reach the result first [1, 3].

The race for the proof

The Navier-Stokes equations, which govern fluid dynamics, have long remained one of the most difficult challenges in mathematical physics [1]. OpenAI’s proof, which suggests the equations can reach "finite-time blowup" where fluid speeds become infinite, was generated by a next-generation model that the company claims outperforms its current GPT-6 Astra system [2, 3]. The process utilized Lean, a software tool that verifies mathematical proofs step-by-step, to ensure the validity of the output [3].

The controversy centers on the timeline of discovery. Buckmaster and Anthropic mathematician Levent Alpöge claim they had been working on a specific, unconventional approach to the problem for nearly a year [1, 3]. Buckmaster states that after he shared information about his progress with an OpenAI mathematician on September 3, he was informed three days later that OpenAI had already produced a 100-page proof using the same narrow methodology [3]. Buckmaster alleges that OpenAI researcher Sébastien Bubeck pressured him to exclude Alpöge from credit and warned him against making the dispute public, citing potential damage to his career [1, 3].

Data usage and competitive stakes

OpenAI denies the allegations, with Bubeck and CEO Sam Altman rejecting the characterization of the events [3]. While OpenAI maintains that no specific user data was accessed to solve the problem, the company acknowledged it cannot rule out that de-identified data derived from researchers' usage of its products—specifically the coding model Codex—may have influenced its model’s training [1, 2].

The dispute highlights the growing friction between academic research and the massive compute resources held by AI labs. OpenAI’s effort consumed 300 billion output tokens, an amount estimated to cost $22.5 million if charged at standard rates [1]. Despite the $1 million bounty offered by the Clay Mathematics Institute for a solution to the Navier-Stokes problem, OpenAI has stated it will not claim the prize [2].

What to watch

• Clay Mathematics Institute verification: Whether the institute formally recognizes the proof as a definitive solution to the Millennium Prize problem.
• Model transparency disclosures: Any further updates from OpenAI regarding how user interactions with models like Codex are utilized in the development of future, more capable systems.
• Academic policy shifts: Potential changes in how mathematicians disclose preliminary findings to AI labs to prevent similar disputes over credit and intellectual property.

The conflict leaves the status of the proof in a state of uncertainty, as the mathematical community evaluates the validity of the work against the backdrop of the intense rivalry between OpenAI and Anthropic. Whether the proof holds up to peer scrutiny remains the primary question, regardless of the ongoing debate over how the result was achieved.