Over the past year, AI labs such as OpenAI and Anthropic announced a series of advances on entrenched mathematical problems, even claiming a solution to one of the seven Millennium Prize problems. While the results would normally be celebrated, the rapid, high-profile releases have provoked sharp criticism from mathematicians who say the announcements have been mishandled.
OpenAI responded by forming an independent advisory group of prominent mathematicians, a move reported by The Verge. The panel is intended to guide OpenAI and other AI firms on how to present and disseminate new mathematical findings. Researchers interviewed by The Verge described the effort as “messy and confusing,” noting that the sudden creation of the group raised doubts about OpenAI’s willingness to change its approach.
The advisory board’s first task, according to OpenAI, is to help coordinate the release of dozens of additional results generated by an unreleased model. Mathematicians expressed concern that a small, elite panel may not reflect the broader community and questioned how much authority the group will actually possess. The lack of clarity about the panel’s decision-making power has left many researchers uneasy about future AI-driven publications.
In a separate announcement, OpenAI claimed its internal model solved the Navier-Stokes equations, one of the Millennium Prize challenges that has remained open for roughly nine decades. The company said the model, described as more powerful than the newly released GPT-6 Astra and run with 10,000 concurrent agents, produced the result after a rapid push to outpace competing researchers. Critics have accused OpenAI of “scooping” and violating long-standing academic norms.
Questions about the provenance of training data have intensified the dispute. Mathematician Andreas Thom posted on Mastodon that earlier interactions with ChatGPT may have seeded the breakthrough, noting that one of OpenAI’s ten announced results relied heavily on his and Gábor Kun’s work on non-sofic groups. Another researcher has publicly labeled OpenAI’s data practices as “unethical” and “dishonest,” demanding greater transparency.
Leading figures such as Fields Medalist James Maynard have voiced personal unease, describing a period of “soul searching” as the discipline confronts AI-generated proofs. The combined effect of rapid breakthroughs, opaque data handling, and a perceived race to claim credit has prompted warnings that the current trajectory could chill mathematical research. OpenAI’s attempts to mend relations through the advisory panel remain under close scrutiny by the community.