Mathematicians Issue ‘Leiden Declaration’ Against Risks of AI in Field
The International Mathematics Federation, in collaboration with dozens of research institutions including the University of Leiden, has released the so-called „Leiden Declaration,” formally titled the Leiden Declaration on AI and Mathematics. The declaration took roughly eight months to draft, originating from a symposium held at Leiden University in September of this year, and has received an endorsement from the International Mathematical Union (IMU). The declaration points out that AI-generated proofs, while seemingly plausible, may contain subtle yet undetectable errors, thereby threatening the verifiability of mathematical results; that AI’s large-scale scraping of mathematicians’ work without attribution is eroding the tradition of academic provenance; and that commercial entities using mathematics to promote AI products are quietly influencing the autonomy of research directions.
The declaration does not call for a ban on AI. Instead, it proposes a three-tier response framework: individual researchers must disclose their use of AI and take responsibility for the correctness of their results; academic institutions and professional organizations should formulate policies upholding standards for peer review and transparency; and policymakers should fund public AI infrastructure and support open science. The declaration specifically highlights that early-career mathematicians and students are particularly vulnerable to these impacts, and that without intervention, the long-term health of the entire discipline will be jeopardized. The declaration has been submitted for inclusion on the agenda of the next International Congress of Mathematicians (ICM), with the hope of sparking broader institutional action.