OpenAI said it began training a new internal model on August 28 and that the system has since solved the Navier–Stokes Millennium Prize problem along with “more than 100 long-standing open problems” spanning most areas of mathematics.¹

The company said the pace caught even its own mathematicians off guard, triggering internal debate over how to alert a field that may have little time to absorb the consequences. OpenAI argues that mathematical discoveries can flow into a broad range of applications, making responsible deployment an issue that reaches well beyond pure research.¹

That claim lands amid a clear warning from parts of the mathematical community. OpenAI cited a recent open letter, A Severe Misalignment of AI in Mathematics, whose authors raised concerns about the negative effects of treating solutions to open problems as benchmarks for increasingly capable systems.¹ The friction is not over whether the tools are powerful, but over who gets to determine what that power does to mathematical work.

In response, OpenAI is working with mathematicians to establish an independent advisory group, hosted at the Institute for Advanced Study. The panel is meant to assess the importance of emerging results, advise on their dissemination and research standards, and recommend ways AI tools could aid research and learning.¹

OpenAI says the group can offer unrequested advice, publicly criticize the company’s impact on mathematics and select its own members; its participants will not be paid by OpenAI.¹ The initial roster includes François Charles, Timothy Gowers, Martin Hairer, Edward Witten and Melanie Matchett Wood, among others.¹

Yet the company has drawn its own line. “The group will not be responsible for advising us on how to pace our internal progress on mathematics,” OpenAI said.¹ Mathematicians are being brought into the conversation—just not the decision over the speed of the race.