The Leiden Declaration on Artificial Intelligence and Mathematics appeared today, following a 2025 workshop at the Lorentz Center in Leiden. It gathers an international cohort of researchers to map how AI tools reshape proof formalisation and other mathematical workflows. The document flags risks tied to automatically generated results and the growing use of proprietary models in academic communities daily practice.

Authors outline four main concerns: reliability of AI‑generated proofs, attribution when proprietary models contribute, and how new tools alter peer review and publication norms. They call for clear disclosure of AI use, rigorous verification by human researchers, and proper citation of pre‑existing work. The declaration also warns that commercial actors could shift the balance of mathematical credit in the global field.

The document offers tailored recommendations. Individual researchers should document AI assistance, confirm results, and cite antecedent findings. Professional bodies and funders are urged to craft AI policies for journals and reviews, preserving rigorous scrutiny. Policymakers must consider regulation, public investment in computational infrastructure, and expert evaluation of AI claims. The declaration serves as a roadmap for safeguarding mathematical integrity today.