International Mathematical Union endorses Leiden Declaration warning of AI threats to mathematics research

The Leiden Declaration on Artificial Intelligence and Mathematics, developed by 16 researchers over eight months and published on June 2, 2026, has been endorsed by the International Mathematical Union. The declaration warns that AI developments are threatening "characteristic values" of mathematical research, often affecting students and early-career mathematicians.

The declaration identifies five specific concerns: AI models producing plausible but unreliable or incorrect arguments difficult to distinguish from correct proofs; AI-generated outputs that do not properly cite human works and are often trained on data obtained through copyright violations; incentivization of AI use that may disadvantage researchers lacking access or unwilling to use proprietary technologies; mathematics achievements communicated through press releases without peer review, leading to media oversimplification; and increasing technology company involvement threatening the autonomy of mathematics as university budgets face pressure.

The declaration points to OpenAI's recent announcement of an AI model disproving an 80-year-old mathematical conjecture as exemplifying concerns about corporate press releases operating on "market timelines before the accepted processes of community evaluation in mathematics can take place." The OpenAI announcement lacked disclosure of prompts, training data, and computational resources used, and the model itself remains proprietary and unavailable outside the company.