Did OpenAI crack a million-dollar math problem using data stolen from mathematicians?

In what has become a contentious issue within the mathematical community, allegations have surfaced that OpenAI may have utilized its powerful computing resources, and possibly academic researchers' private data, to solve a highly complex mathematical problem. The incident has sent shockwaves through the science world, with prominent figures like NYU mathematician Tristan Buckmaster comparing it to an historic moment in chess.

The Navier-Stokes equations, which describe the motion of viscous fluids, have baffled mathematicians for over two centuries and were designated as one of the Millennium Prize Problems by the Clay Mathematics Institute, offering a $1 million prize for a solution. The problem involves whether these equations always produce smooth, predictable solutions or if they can experience a sudden, uncontrolled blow-up.

Spanish mathematicians Diego Córdoba and Luis Martínez-Zoroa have been working on a specific approach to solving this problem, focusing on creating blow-ups through small, concentrated changes in fluid motion. Recent developments indicate that Buckmaster and Alpöge, using large language models like Anthropic's Claude and OpenAI's GPT-5.6 Sol, have made significant progress in proving blow-ups for modified versions of the Navier-Stokes equations.

However, the raw output from the AI was difficult to decipher, prompting the pair to spend weeks refining it into a conventional mathematical paper. Rumors emerged in mid-September suggesting that OpenAI had indeed cracked the problem, but Buckmaster preemptively alerted a colleague to verify the claims before publication.