A Scientific Breakthrough Or Data Misuse
OpenAI announced this week that its new internal artificial intelligence model successfully solved a major aspect of the Navier-Stokes equations in 88 hours. These equations describe fluid motion and have challenged mathematicians for nine decades. The company assigned 10,000 autonomous AI agents to the project to crack the proof. By the end of the process, the agents exchanged nearly three million messages and consumed 130 billion output tokens to reach their conclusion. OpenAI representatives noted the cost of this computing effort would reach roughly $10 million at current market rates.
The Navier-Stokes existence and smoothness problem remains one of the most difficult challenges in physics and mathematics. Understanding turbulence requires mastery of these specific equations. While OpenAI claims its result is a milestone for machine intelligence, the firm stated it does not plan to claim the $1 million Millennium Prize offered by the Clay Mathematics Institute for a complete proof. The current solution addresses two of the four requirements specified by the prize criteria. The findings have not yet undergone formal peer review or verification by external experts.
Allegations of Intellectual Property Missteps
Controversy emerged immediately following the announcement. Tristan Buckmaster, a professor at New York University, alleged that OpenAI obtained details about his own research on the same problem before releasing their findings. Buckmaster claims he and Levent Alpöge, a researcher at Anthropic, were using OpenAI tools to work on the Navier-Stokes challenge. He stated that he discovered information regarding his progress had reached OpenAI personnel on September 3, hours before the company published its results.
Buckmaster provided email records to support his claim that OpenAI lacked interest in the problem until learning of his progress. He characterized the company's behavior as an attempt to take credit for work that was already underway by independent researchers. The professor expressed that he felt forced to speak out to clarify the timeline of events. He has not yet reviewed the full technical proof offered by OpenAI but noted the ethical implications of the company’s internal process are significant.
The Technical Response from OpenAI
OpenAI representatives countered these accusations by stating they had no access to the work conducted by Buckmaster and Alpöge. The company acknowledged that they heard rumors about Millennium Prize problems being solved on September 1, which prompted them to deploy their internal model against the challenge. They suggested that while they cannot rule out that de-identified user data may have influenced the training of their models, their final proof differs from the one produced by the external mathematicians.
OpenAI praised the efforts of the two researchers as remarkable. The company stated its internal model is significantly more capable than any public release currently available. This model remains restricted to internal use due to its performance levels. The broader significance of the event remains unclear as the industry watches for independent verification of the proof. If confirmed, this would represent the first time a major mathematical mystery yielded to an AI-driven approach, marking a shift in how theoretical research is conducted.