A group of prominent mathematicians unveiled the Leiden Declaration on Artificial Intelligence and Mathematics Monday, marking the most coordinated response yet from an academic discipline to the way AI firms are exploiting scholarly work. Endorsed by the International Mathematical Union and signed by Fields Medal recipient Peter Scholze, the declaration condemns the practice of training AI models on published papers without the authors’ permission and of releasing breakthrough results through press releases rather than peer‑reviewed journals.
The document identifies five specific threats AI poses to the discipline. First, current systems generate plausible but unreliable arguments that can masquerade as valid proofs, making it hard for researchers to separate genuine results from subtle errors. Second, AI‑generated content often omits proper citation, effectively erasing the contributions of human mathematicians. Third, the lure of AI incentives is distorting hiring, funding and recognition within the field. Fourth, the rush to publicize AI‑driven discoveries—exemplified by Google DeepMind’s AlphaProof, which solved three International Mathematical Olympiad problems in 2024 but delayed peer‑review publication for over a year—undermines the traditional vetting process. Fifth, research agendas risk being steered toward problems that are easy for machines to tackle, sidelining questions valued for their intrinsic significance.
To counter these risks, the declaration offers recommendations at four levels. Individual researchers are urged to disclose any AI tools used in their work, retain full responsibility for the correctness of their results, and refuse to list AI systems as co‑authors. Academic societies should demand that AI‑derived findings meet rigorous standards, negotiate licensing agreements that prevent unauthorized use of published material for training, and insist on peer‑reviewed dissemination. Policymakers are asked to increase public oversight of AI companies and invest in non‑proprietary computational infrastructure, echoing European regulatory models.
Signatories span a broad spectrum of the mathematical community. In addition to Scholze, the declaration bears the names of Robbert Dijkgraaf, former Dutch minister of education and president‑elect of the International Science Council, and Steven Strogatz of Cornell University. Imperial College’s Kevin Buzzard, a leading advocate for formalized mathematics, praised the statement as a “well‑thought‑through response” to the ongoing disruption.
The authors also confront AI firms directly, warning that the commercial drive to showcase AI capabilities can lead to overstated claims and the deployment of mathematical models in ethically fraught applications such as warfare, mass surveillance and the erosion of democratic processes. By insisting that mathematicians retain control over how their work is used, the Leiden Declaration seeks to preserve the discipline’s human‑centered ethos while acknowledging the potential benefits of AI when applied responsibly.
Developed over eight months by a 17‑member working group after a 2025 workshop at the Lorentz Center in Leiden, the declaration opened with 37 verified signatures and invites additional endorsements from the global mathematical community. Its release signals a turning point: mathematicians are no longer passive observers of AI’s rise but active architects of the standards that will shape the future interaction between human insight and machine intelligence.
This article was written with the assistance of AI.
News Factory SEO helps you automate news content for your site.