OpenAI Claims Math Milestone Amid Data Integrity Questions
AI-generated image · Bay Street Wire
The company says it solved the 90-year-old Navier-Stokes problem, but researchers suggest the breakthrough may have relied on their own proprietary drafts.
OpenAI has announced a solution to the Navier-Stokes problem, a complex mathematical challenge involving the flow of gas and liquid that has remained unsolved for approximately 90 years. As reported by The Verge, OpenAI utilized 10,000 concurrent agents and an internal AI model—which the company describes as more powerful than GPT-6 Astra—to reach the solution. The internal model began training on August 28.
While the Navier-Stokes problem is one of seven Millennium Prize Problems carrying a $1 million reward, OpenAI stated it does not intend to claim the prize. However, the announcement has sparked immediate controversy. According to The Verge, Tristan Buckmaster, a mathematics professor at New York University, and Levent Alpöge, a researcher at Anthropic, published findings on a related problem just one day prior to OpenAI's announcement.
Buckmaster has raised concerns regarding whether OpenAI accessed his and Alpöge's project drafts, which were entered into OpenAI's Codex and Anthropic's Claude. In a statement, Buckmaster questioned if the model was trained on these sessions; he claims that while he was told the model did not look up user data, he received no answer regarding training.
OpenAI has denied accessing specific user data to solve the problem, though the company admitted it cannot rule out that "de-identified data" from product usage helped improve its models. Sebastien Bubeck, a member of technical staff at OpenAI, told The Verge that the company did not see the work of Buckmaster and Alpöge until it was released publicly and asserted that the resulting proofs differ significantly.