OpenAI announced on Sept. 9 that an internal AI model had made progress toward solving the Navier-Stokes existence and smoothness problem, one of the Clay Mathematics Institute’s seven Millennium Prize Problems. The model, described as significantly more capable than GPT-6 Astra, allegedly addressed a centuries-old question in fluid dynamics, which governs phenomena like weather patterns, ocean currents, and blood flow.
Mathematician Tristan Buckmaster of New York University (NYU) publicly raised concerns about OpenAI’s claims, stating that the company’s internal model may have been trained on his and Levent Alpöge’s (an Anthropic employee) unpublished research. Buckmaster, who collaborated with Alpöge using AI tools from both OpenAI and Anthropic, did not confirm whether OpenAI accessed their specific work. OpenAI responded that no user data was directly accessed but could not rule out the use of de-identified data derived from broader model training.
The dispute centers on scientific credit, AI training practices, and the transparency of AI-assisted discoveries. Buckmaster emphasized the need for unhurried discussion about the implications of AI’s role in mathematical breakthroughs, comparing the moment to the 1997 chess match where IBM’s Deep Blue defeated Garry Kasparov. Meanwhile, OpenAI framed its achievement as a milestone in AI’s capability to tackle complex scientific problems.
What is the Navier-Stokes problem?
The Navier-Stokes equations describe the motion of fluids and have been a cornerstone of physics and engineering since the 19th century. The existence and smoothness problem—one of the seven Millennium Prize Problems—asks whether solutions to these equations always remain well-behaved or can break down unpredictably. A solution would not only earn a $1 million prize but also revolutionize fields like aerodynamics and climate modeling.
OpenAI’s Claim vs. Mathematicians’ Concerns
OpenAI’s Sept. 9 announcement stated that its internal model had made measurable progress toward solving the problem. The company did not claim a complete solution but described the development as a significant step in AI’s mathematical reasoning capabilities. OpenAI spokesperson Kayla Wood stated, “This represents a leap in how AI can assist in solving long-standing scientific challenges.”
Buckmaster and Alpöge’s work, shared in a four-page statement and accompanying papers, outlined new insights into the problem but did not claim a full solution. Buckmaster alleged that OpenAI’s model may have been influenced by their research, though he acknowledged uncertainty about whether specific data was used. OpenAI countered that no user data was accessed, though it acknowledged that de-identified training data could have contributed to the model’s performance.
Broader Implications for AI and Science
The episode has intensified debates over AI’s role in scientific discovery, particularly regarding data sourcing and attribution. The incident also raises questions about how AI models are trained and whether their outputs should be subject to peer review before being touted as breakthroughs. Mathematician Terence Tao, a Fields Medalist, called Buckmaster and Alpöge’s work “remarkable” in a Mastodon post, but did not endorse OpenAI’s claims.
Industry and Academic Reactions
The dispute has drawn attention from both the AI and mathematics communities. Some researchers praised OpenAI’s progress as evidence of AI’s potential to accelerate scientific discovery, while others criticized the lack of transparency. The Clay Mathematics Institute, which administers the Millennium Prize, has not commented on the claims.
What’s Next?
OpenAI has not provided detailed evidence of its model’s solution, and Buckmaster has called for further scrutiny of how AI models are developed and credited. The incident underscores the need for clearer standards in AI-assisted research, particularly as models grow more advanced. Mathematicians and AI developers are expected to engage in broader discussions about collaboration, data usage, and the ethical implications of AI in scientific breakthroughs.