AI and Mathematics: What the Leiden Declaration Says

What does AI progress in mathematics mean?

Mathematics depends on arguments that can be examined step by step. A proof is a reasoned demonstration that a statement follows from accepted assumptions and earlier results. AI tools can assist with parts of mathematical research, and some systems have produced work described as research-level mathematics. ¹

That does not mean every AI-generated answer is correct, new or adequately explained. A result still needs scrutiny: readers need to understand the reasoning, check whether it holds, and see how it relates to earlier work. The central issue is not simply whether a machine can produce a promising answer, but whether people can verify and build on it.

What is the Leiden Declaration?

The Leiden Declaration on Artificial Intelligence and Mathematics is a statement about how the mathematical community should respond to AI’s growing role in research. It began with a workshop at the Lorentz Center at Leiden University in September 2025. An international group of mathematicians published the declaration in June 2026, and the International Mathematical Union endorsed it. ¹

The declaration was written by 16 mathematicians, including Jim Portegies, who led the project, Mateja Jamnik, Ursula Martin, Rodrigo Ochigame and Michael Harris. Its purpose is to make the values of mathematics explicit as AI becomes part of research. ¹

The authors call for disclosure when AI is used in research papers, proper peer review, and ways for results to be checked independently. Peer review is the process in which other experts assess research before it is accepted for publication. The declaration also argues that human contributions and earlier work should receive appropriate credit. ¹

Why do mathematicians see both promise and risk?

AI may help researchers explore difficult problems or develop arguments. But a generated proof can be hard to assess if the methods are not explained or if important steps cannot be checked. The declaration warns that overreliance on AI-generated proofs could weaken the accuracy and reliability of mathematical research. ¹

There are concerns beyond correctness. The declaration says that companies’ involvement could steer attention toward questions that AI tools handle well, rather than questions chosen for their broader importance. Researchers without access to those tools could also be left at a disadvantage. ¹

Giving credit is another concern. The declaration notes that models trained on published work may produce material without properly acknowledging the human work behind it. It also raises questions about the use of copyrighted research in training data. ¹

What does this mean for an ordinary reader?

For readers, the practical point is that an AI-assisted mathematical result should not be treated as confirmed just because a system or company presents it as a discovery. The declaration’s proposed safeguards—clear disclosure, expert review and independent verification—are ways to make it easier to judge how a result was reached and whether it holds. ¹

This debate is about the standards that make mathematical knowledge trustworthy, not a claim that AI has replaced mathematicians. The declaration calls for the community to preserve openness, honest attribution and access to methods while deciding how these tools should fit into research. ¹

The available source describes the declaration and the concerns behind it, but does not establish how widely its recommendations have been adopted. Readers looking for more detail can consult the declaration itself and the International Mathematical Union, which endorsed it. ¹