Artificial intelligence is moving deeper into mathematics, and that shift is now drawing a sharper response from leading researchers. Twenty-five prominent mathematicians, each a Fields Medal recipient, have signed an open letter urging AI labs to protect attribution, transparency, and the long-term culture of open research.
The discussion intensified after NYU professor Tristan Buckmaster said OpenAI pressured him not to credit a collaborator from Anthropic in connection with a major math problem. He also questioned whether OpenAI's work with Codex helped shape its recently announced proof related to the Navier-Stokes equations.
OpenAI later withdrew its sponsorship of a math event at Caltech after criticism from researchers there. The broader debate, however, goes beyond one company or one proof: mathematicians are asking how AI-generated results should be shared, verified, and integrated into the field.
The signatories argue that speed alone is not enough. In their view, a mathematical breakthrough only becomes meaningful when it can be explained, checked, and built upon by the community. They warn that rushed announcements can leave little room for proper documentation, recognition of prior work, and the careful transfer of ideas that keeps the discipline evolving.
The letter also echoes the Leiden Declaration, a June framework that outlined recommendations for mathematicians, institutions, and policymakers as large language models become more capable of producing proofs. Together, these efforts point to a larger question for science and creative work: how to use AI without weakening the human systems that give discoveries lasting value.
As AI continues to expand across research, the future may belong to teams that combine machine speed with human judgment, clarity, and shared knowledge.
