A group of leading mathematicians has issued an unusual collective warning about the growing race to use difficult mathematics problems as benchmarks for artificial intelligence. Terence Tao, a 2006 Fields Medal winner and UCLA professor, said the group had discussed the issue over the previous week and decided the concern was urgent enough to publish a statement quickly.
What the mathematicians object to
The initial signatories include 25 Fields Medal recipients, among them Princeton professor June Huh and Peter Scholze of the Max Planck Institute for Mathematics. The statement acknowledges rapid improvements in AI's mathematical abilities but argues that technology companies are increasingly treating hard mathematics problems as competitive performance tests in ways that do not align with the goals of the mathematical research community.
The concern is not simply that AI can solve harder problems. The signatories argue that benchmark-driven competition can distort incentives around how research questions are selected, shared and valued, potentially damaging the academic ecosystem that produces those problems in the first place. Their intervention adds a high-profile disciplinary response to the wider debate over how AI capability races should interact with scientific communities and research norms.