OpenAI Navier-Stokes Solution Controversy

OpenAI Claims Solution to Millennium Prize Problem

OpenAI has published a full proof of the Navier-Stokes existence and smoothness problem, one of the seven Millennium Prize problems. The company reports that the solution was discovered by an unreleased next-generation model in a week-long effort that consumed 300 billion output tokens, representing approximately $22.5 million in compute costs based on current Astra rates.

The Navier-Stokes equations are fundamental to fluid mechanics, but their theoretical existence and smoothness have remained unsolved. A formal proof represents a significant advancement in mathematical physics and carries a $1 million bounty from the Clay Mathematics Institute.

Allegations of Research Scooping and "Dirty" Tactics

NYU mathematics professor Tristan Buckmaster, who collaborated with Anthropic mathematician Levent Alpöge, alleges that OpenAI intentionally scooped their work after obtaining information about their progress. Buckmaster and Alpöge had been working on a preliminary finding using a combination of Codex and Claude AI models.

Key points of the dispute include:

• Timing and Approach: Buckmaster asserts that the specific mathematical route they chose (options c and d in Fefferman's statement of the problem) is uncommon. He argues that OpenAI's sudden pivot to the same approach on September 1—after rumors of a solution had circulated—is highly suspicious, stating, "It is not the direction one arrives at in a few days by giving a model the problem statement."
• Compute Advantage: Buckmaster suggests that OpenAI, upon realizing the research direction was viable, leveraged its massive computing resources to reach a formal proof before the original researchers could publish.
• Professional Conduct: Buckmaster alleges that during disputes over credit, OpenAI's Sebastien Bubeck suggested removing Levent Alpöge's credit (due to Alpöge's affiliation with Anthropic) and reportedly told Buckmaster, "Why would you ruin your career?" and "If you don’t want me to be nice, then I don’t have to be nice."

Data Privacy and Model Regurgitation

Because Buckmaster used OpenAI's Codex extensively during his research, he raised concerns that his work may have been absorbed into OpenAI's training data. OpenAI reserves the right to train models on Codex interactions unless users opt out.

OpenAI has denied accessing specific user data to solve the problem, stating:

"We (the researchers and the agents) did not see any of their work through any means until they released it publicly — in particular, no specific user data was accessed in order to solve this problem."

However, the company conceded that it cannot rule out that "de-identified data derived from their usage of our products helped improve our models." OpenAI further argues that their proof differs significantly from Buckmaster's, specifically regarding the Euler case (forced vs unforced).

Community Analysis and Implications

The incident has sparked significant debate within the technical and academic communities regarding the ethics of AI-assisted research:

• The "Compute Scoop": Commentators have noted that this represents a new form of academic rivalry where a lab with superior compute can effectively "out-compute" a human researcher's insight once the direction is known.
• Academic Ownership: Some argue that if a model was trained on a researcher's private interactions to reach a conclusion, the academic ownership should remain with the researcher regardless of the legal terms of service.
• AI Collaboration: Despite the controversy, some observers view the event as an "existence proof" that human-AI collaboration can solve some of the most difficult problems in theoretical mathematics.
• Economic Viability: The cost of the discovery ($22.5 million) far exceeds the Millennium Prize bounty ($1 million), suggesting that OpenAI's motivation was likely based on demonstrating AGI capabilities rather than financial gain.