A community-driven initiative addressing artificial intelligence's role in mathematical research has gained backing from 874 mathematicians, including Fields Medal winners Peter Scholze and Terence Tao. The Leiden Declaration on Artificial Intelligence and Mathematics, published June 2, 2026, establishes guidelines for AI integration while preserving mathematics as "a profoundly human endeavour," according to the International Mathematical Union's affirmation.
Declaration Origins and Development
The declaration emerged from a September 2025 conference at the Lorentz Center in Leiden, Netherlands. A working group spent eight months refining recommendations before publication, resulting in a document that addresses five core values and five critical threats facing mathematical research in the AI era.
Notable signatories include:
- Peter Scholze (Max Planck Institute for Mathematics, Fields Medal winner)
- Terence Tao (UCLA, Fields Medal winner)
- Robbert Dijkgraaf (University of Amsterdam, President-Elect of International Science Council)
- Jeremy Avigad (Carnegie Mellon University)
- Kevin Buzzard (Imperial College London)
Five Characteristic Values Mathematics Must Preserve
The declaration identifies proof as the foundation of mathematical certainty, emphasizing clear authorship with accountability for correctness. It calls for transparency and independent verifiability of arguments, shared standards for evaluating mathematical work, and mathematician autonomy in shaping research directions.
Critical Threats from AI Integration
The declaration identifies five major concerns: AI-generated plausible but unreliable arguments that are difficult to distinguish from correct proofs, inadequate attribution when models synthesize published works, skewed research incentives favoring AI-amenable problems, premature results announcements bypassing peer review, and technology company influence threatening research autonomy.
Actionable Recommendations for the Community
The document provides guidance for mathematicians, organizations, policymakers, and industry. Key recommendations include tool disclosure requirements, attribution rigor, peer-reviewed publication standards, and adherence to open science principles. The International Mathematical Union's endorsement adds significant institutional authority to these guidelines.
Key Takeaways
- The Leiden Declaration on AI and Mathematics has 874 signatories, including Fields Medal winners Peter Scholze and Terence Tao
- Published June 2, 2026 with DOI 10.5281/zenodo.20302944 following an eight-month development process
- Identifies five critical threats including AI-generated unreliable proofs and inadequate attribution of synthesized works
- The International Mathematical Union affirms mathematics "must be guided by human judgment, fair and transparent practices"
- Provides actionable guidance for mathematicians, organizations, policymakers, and industry on responsible AI integration