OpenAI announced on Tuesday that one of its unreleased AI models had solved the Navier-Stokes problem, a fluid dynamics puzzle that is one of mathematics' seven Millennium Prize problems and had defeated human researchers for nearly 90 years. Rather than being celebrated outright, the achievement has been overshadowed by accusations that OpenAI rushed to beat rival academics to the answer after learning of their progress, with critics saying the company's conduct breached long-standing norms of academic collaboration and trust.

OpenAI said an internal model took just 88 hours to crack the problem using around 10,000 AI agents, an effort it called a "milestone" despite reportedly costing millions of dollars, though it does not intend to claim the $1 million prize. The controversy stems from NYU professor Tristan Buckmaster, who published related findings with an Anthropic researcher one day before OpenAI's announcement; Buckmaster says an OpenAI researcher responded to his questions with hostility, including warning him that going public could "ruin his career," and urged him to credit OpenAI's model instead. OpenAI denies accessing his work directly but admits it cannot fully rule out that de-identified user data indirectly influenced its model, a response that mathematicians such as Queen Mary University's Abhishek Saha say reflects behaviour "mathematicians will generally not do."