25 Fields Medalists accuse OpenAI of harming mathematics community

25 recipients of the Fields Medal, mathematics' highest honor, have signed an open letter accusing artificial-intelligence laboratories of threatening the discipline's intellectual work as they race one another to solve famous open problems. The signatories argue that any benefit from resolving such challenges will materialize only if the solutions are comprehensible and communicable to the mathematical community, and ultimately to the wider world.

The dispute escalated this week when Tristan Buckmaster, a professor at NYU, alleged that OpenAI pressured him not to credit a research collaborator at Anthropic who had co-solved an important problem. Buckmaster questioned whether OpenAI had used their joint work with Codex to generate its own purported breakthrough proof during an entire weekend of inference. OpenAI's proof remains unverified.

On Thursday OpenAI withdrew its sponsorship of a mathematics event at CalTech after criticism from researchers at the institute. The letter notes that announced solutions are frequently rushed, leaving no time for careful write-up, for isolating new methods and ideas, or for citing relevant prior work. This, the mathematicians say, raises serious questions of attribution and intellectual appropriation. They warn that without mathematicians to nurture those ideas and weave them into the canon, concepts developed by AI will never fully "come alive," and the human chain of transmission will be lost.

A genuine chilling effect is emerging: other researchers are growing paranoid, wondering whether their own use of Codex has been fed back into OpenAI's newer models. The culture of open research is under threat; if frontier labs identify a promising path to a discovery, they can spend tens of millions of dollars running LLM to beat the original researchers to the proof — a dynamic that incentivizes secrecy. The letter follows the "Leiden Declaration," published in June by a working group of mathematicians, which offered recommendations for researchers, institutions, and policymakers.

As in software engineering and other fields where AI tools are reshaping workflows, the mathematicians locate the core value in the "work around the work": the worth of mathematics lies not only in proofs and the question of credit, but in the intellectual superstructure that trains students, identifies new questions and ideas, and integrates them into the broader human civilization. The challenges the mathematical community now faces mirror those of other scientific and creative professions, the signatories write, and point to difficulties all of humanity may confront: how to ensure that when AI changes the way work is done, we do not lose sight of what the work was meant to achieve in the first place.