Overview
- The Leiden Declaration was published on Tuesday, June 2, 2026, and the International Mathematical Union publicly endorsed the 11‑page statement prepared by a 16‑person working group after a September 2025 Leiden workshop.
- The declaration warns that current commercial AI tools can produce convincing but unreliable proofs that are hard to verify, a problem that strains journal peer review and risks polluting the literature with incorrect results.
- Authors call for mandatory disclosure of AI use in research, retention of human responsibility for correctness, proper attribution when models draw on published work, and stronger peer‑review standards for AI‑assisted results.
- The statement criticizes tech companies for announcing mathematical claims on market-driven timelines without sharing prompts, training data, or compute details, and it urges legal measures and public computing infrastructure to level access between academia and industry.
- Signatories reported range from more than 130 to hundreds of names, the declaration spotlights risks to students and early‑career researchers, and its recommendations will be discussed at upcoming IMU meetings and the International Congress of Mathematicians in Philadelphia.