TL;DR
The Leiden Declaration on Artificial Intelligence and Mathematics, endorsed by the International Mathematical Union and signed by Fields Medal recipient Peter Scholze, calls on mathematicians to confront how AI companies are using published research without consent, bypassing peer review, and threatening the integrity of proof and attribution.
A coalition of mathematicians from institutions including Oxford, Cambridge, ETH Zurich, Columbia, and Northwestern has published a formal declaration calling on the mathematical community to confront the threats that artificial intelligence poses to their discipline. The Leiden Declaration on Artificial Intelligence and Mathematics, released on Monday and endorsed by the International Mathematical Union, is the most significant collective response from a major academic discipline to the way AI companies are using, and in some cases exploiting, published research.
The 11-page document does not oppose AI in mathematics. It opposes the way AI companies are treating mathematical work: training models on published papers without consent, announcing results through press releases instead of peer review, undermining attribution, and reshaping research priorities to serve commercial interests rather than intellectual significance. “Mathematics is, and should always remain, a profoundly human endeavour,” said Ulrike Tillmann, vice president of the IMU.
Five threats to mathematical research
The Declaration identifies five specific ways AI threatens the values that make mathematics trustworthy. First, current AI systems produce plausible but unreliable arguments that are difficult to distinguish from correct proofs. This applies not only to informal reasoning but also to formal computer-encoded proofs, where the difficulty lies in translating between machine and human representations of concepts. The problem of AI-generated content that looks authoritative but contains subtle errors is not unique to mathematics, but in a discipline built on certainty, it is existential.
Second, AI models trained on published mathematical work do not properly cite the human contributions they synthesise. The Declaration notes that much training data was obtained by “systematically exploiting licences and access arrangements that were not made with artificial intelligence in mind, or indeed by simply violating copyright protections.”
Third, the use of AI is becoming incentivised for its own sake, distorting hiring, funding, and recognition. Fourth, results are increasingly communicated through press releases and blog posts rather than peer-reviewed journals, seeking publicity “on market timelines before the accepted processes of community evaluation in mathematics can take place.” The Declaration cites Google DeepMind’s AlphaProof, which solved three International Mathematical Olympiad problems in 2024 but took more than a year to publish its methods in a peer-reviewed venue. Google’s broader AI strategy relies on mathematical reasoning capabilities as evidence of general intelligence, creating commercial incentives to announce results before the mathematical community can properly evaluate them.
Fifth, the autonomy of mathematics is under threat. Research questions may come to be prioritised because they are amenable to automation rather than because experts judge them to be deeply significant. “Indeed, broader understanding of the field may be permanently lost in the process of automation,” the Declaration warns.
What it recommends
The Declaration makes recommendations at four levels. Individual mathematicians should disclose all AI tool use in papers, retain personal responsibility for the correctness of results, refuse to grant authorship to AI systems, and “consider carefully which tools to use” based on whether their developers align with the Declaration’s values.
Mathematical organisations should insist that results obtained by automated techniques meet standards that address the specific risks those techniques introduce, protect authors’ rights by developing licensing agreements that prevent use of published work as training data without consent, and demand that results continue to be published through peer-reviewed venues. European regulatory frameworks provide a model, but the Declaration argues that the mathematical community must also set its own standards independently of government.
For policymakers, the recommendations are blunt. “Don’t believe the hype,” the Declaration states. “There is currently a strong commercial incentive on the part of the technology industry to overstate the capabilities of their products.” It calls for significantly increased public oversight of the AI industry and investment in public computational infrastructure as an alternative to proprietary systems.
Who signed it
The Declaration carries significant weight because of its signatories. Peter Scholze, a Fields Medal recipient and director of the Max Planck Institute for Mathematics, endorsed it with a personal statement: “I am pondering my mathematical ideas without use of AI, and generally avoid reading AI-generated text as best as I can.” Other endorsements came from Robbert Dijkgraaf, former Dutch minister of education and president-elect of the International Science Council, and Steven Strogatz, Cornell’s distinguished professor for the public understanding of science and mathematics.
Kevin Buzzard, the Imperial College professor who has been one of the most prominent advocates for formalised mathematics, called it “a well-thought-through response to what is currently happening, as AI continues to disrupt this space.” The tension between AI capability and research integrity that the Declaration describes is not limited to mathematics, but mathematicians are among the first academic communities to respond with a coordinated, institution-backed statement.
The deeper argument
The Declaration’s most provocative section addresses AI companies directly. It argues that tech companies are attracted to mathematics because formalised proofs can be checked automatically, creating an “effectively unlimited source of feedback for training artificial intelligence models.” The strategy rests on an assumption that capabilities developed through mathematical theorem proving will extend to broader general reasoning, an assumption the Declaration treats sceptically.
“Some of the resulting general-purpose models are being commercialised for applications that raise grave ethical concerns,” the authors write, “including warfare, oppression, mass surveillance, and the undermining of democracy.” The intersection of AI research and military applications has become one of the defining tensions of 2026, and the Leiden Declaration makes clear that mathematicians do not want their work used as training data for systems deployed in those contexts without their consent.
The Declaration was developed over eight months by a 17-member working group following a September 2025 workshop at the Lorentz Center in Leiden. It had 37 verified signatories on its first day and is open for additional signatures from the mathematical community.
