The Leiden Declaration Sets Rules for AI in Mathematics

The Leiden Declaration on Artificial Intelligence and Mathematics asks the field to defend proof, human authorship and peer review as AI moves into research.

Published on June 2, 2026 and endorsed by the International Mathematical Union, it grew out of a Lorentz Center conference in Leiden attended by around 60 researchers from mathematics, computer science, philosophy and history.

More than 3,600 people have signed, among them Terence Tao of UCLA, Peter Scholze of the Max Planck Institute and Robbert Dijkgraaf of the University of Amsterdam.

The text names five threats: automated systems that produce plausible but incorrect proofs, models that fail to credit their sources, incentives that reward capability over significance, results announced by press release instead of peer review, and corporate money steering what gets studied.

Individual mathematicians are asked to disclose which tools they used, keep responsibility for correctness, and reserve authorship for people.

Organizations are urged to set publishing and review policy and fund independent university labs, policymakers to regulate industry practice and invest in public computing, and commercial developers to protect employee freedom of conscience and keep their systems out of warfare, oppression and mass surveillance.

As IMU vice president Ulrike Tillmann puts it, mathematics is and should always remain a profoundly human endeavour.

The declaration's quiet power is who signed it after the famous names: the count stands at 3,659 and climbs daily, which turns a workshop of sixty into a referendum. What mathematicians are defending, Fields medalist Peter Scholze among them, is human understanding rather than verified output, and the pressure point is economic, since proofs are becoming nearly free to generate while remaining expensive to check, exactly the imbalance Oxford computer scientist Leslie Ann Goldberg warns will clutter the literature with wrong results built upon, per Leiden University. Communications of the ACM argues every discipline faces the same reckoning next. Watch whether journal publishers convert these principles into required disclosure policies, because norms written for proofs tend to travel.