Twenty-five Fields Medalists have signed an open letter warning that AI companiesā€™ race to solve famous mathematical problems risks undermining the research community that gives those results meaning. The declaration, ā€œA Severe Misalignment of AI in Mathematics,ā€ was reported by TechCrunch on September 11.

Signatories listed on the letter include Terence Tao, Peter Scholze, Maryna Viazovska and June Huh. Their concern is not simply whether a model can produce a correct answer. They argue that solving a problem is a means toward conceptual understanding, new methods and the education of future researchers.

Beyond the benchmark

The letter says rushed announcements can leave too little time for proper writeups, explanations of new ideas and acknowledgment of previous work. It raises attribution and plagiarism concerns, while emphasizing the human effort needed to develop results into material that other mathematicians can understand, teach and build upon.

That distinction matters for evaluating research-oriented AI: a solved-problem count does not, by itself, capture whether a result has been explained or integrated into the fieldā€™s knowledge. The authors describe a risk that rapidly producing answers could weaken the processes that cultivate students and new questions.

A call to change priorities

The declaration also acknowledges AIā€™s potential to accelerate genuine mathematical study. It calls for urgent discussion among mathematicians, technology companies and society about how that potential is directed.

This is a collective statement about research priorities and incentives, not a technical evaluation of a particular model or a verification of any specific AI-generated proof.