26 Fields Medal winners warn that AI's push into mathematics is misaligned

The mathematics community has published a declaration warning that the way AI companies are driving large language models into mathematical research is harmful to the field itself. Titled "A Severe Misalignment of AI in Mathematics", the statement is hosted at mathandai.org and carries the DOI 10.5281/zenodo.22737750.

A statement signed by Fields medalists

The declaration is signed by 26 winners of the Fields Medal, mathematics' most prestigious award, among them Artur Avila, Manjul Bhargava, Pierre Deligne, Simon Donaldson, Vladimir Drinfeld, Martin Hairer, June Huh, Maxim Kontsevich, James Maynard, Shigefumi Mori, Ngô Bảo Châu, Andrei Okounkov, Peter Scholze, Terence Tao, Maryna Viazovska, Cédric Villani and Efim Zelmanov.

The signatories acknowledge that over the last few months the mathematical capabilities of LLMs have improved dramatically, to the point where they can solve major outstanding problems in many fields of mathematics. Their objection is not to progress, but to how that progress is being steered.

Solving problems is a proxy, not the goal

Research mathematics, the document argues, is about understanding the basic structures of shapes, numbers and natural phenomena. Famous problems have served as landmarks against which the community measures its understanding; solving one was always a sign of new insights and methods, which other mathematicians then study through talks, discussions and simplifications.

The declaration warns that the rapid mass production of "true/false" statements could destroy the fertile ground from which new ideas grow. Solutions are often announced in a rush, leaving no time for a proper write-up, the isolation of new methods or citations of relevant earlier work — which raises serious attribution and plagiarism questions. The authors add that without mathematicians willing to develop and integrate AI-conceived ideas into the canon, those ideas may never become fully alive, and the human transmission chain between researchers would be lost.

A wider warning about intellectual work

The signatories place the problem in a broader context: misalignment between the outcomes of AI use and its original purpose affects other scientific and creative professions too, and points to questions all of humanity may face as AI changes how work is done.

The document does not reject AI outright. It states that AI offers the potential to enhance and accelerate genuine mathematical study. Whether that happens will be determined largely by the decisions of the humans in control of the technology — and the issues, it says, must be addressed urgently by the mathematical community, by the companies developing these systems and by society at large.