Sixteen mathematicians have published the 'Leiden Declaration,' warning of potential threats to mathematics, including the accuracy and reliability of AI.

A working group of 16 researchers from 15 universities published the 'Leiden Declaration on Artificial Intelligence and Mathematics' on June 2, 2026, citing concerns that the use of AI in mathematical research could affect the reliability of proofs, copyright and attribution of contributions, and the autonomy of research. The declaration does not aim to eliminate AI from mathematics, but rather calls for the establishment of norms regarding transparency, peer review, human responsibility, and relationships with industry in using AI while protecting the core values of mathematics.

Leiden Declaration on Artificial Intelligence and Mathematics
https://leidendeclaration.ai/

Leiden Declaration: AI is challenging the core values of mathematics - Leiden University
https://www.universiteitleiden.nl/en/news/2026/06/leiden-declaration-warns-ai-is-challenging-the-core-values-of-mathematics

The Leiden Manifesto raises the question of who is responsible for errors and for evaluating correct results when mathematical proofs are created not by humans, but by algorithms whose mechanisms are not fully understood. It also states that it is no longer a hypothetical question to distinguish whether a proof generated by AI is truly new or merely a rearrangement of existing research without proper citation.

This declaration originated from a workshop called 'Mechanization and Mathematical Research' held at the Lorenz Center at Leiden University in the Netherlands in September 2025. Approximately 60 mathematicians, computer scientists, humanities and social science researchers, and policymakers from 10 countries participated in the meeting, and a 16-person working group convened by Jim Portegies of Eindhoven University of Technology subsequently drafted the declaration.

The declaration takes the stance that mathematics is not merely an accumulation of results, but a human endeavor that fosters understanding, clarity, and judgment. It then lists the following as values of mathematical research: 'high certainty through proof,' 'author's responsibility for results,' 'verifiability by third parties,' 'community standards for evaluating the depth and importance of research,' and 'the ability to autonomously shape research topics.'

The declaration further pointed out that AI poses five threats.

◆First threat: An increase in plausible but false proofs and arguments.
The declaration warns that current automation technologies can generate inaccurate arguments that are difficult to distinguish from correct mathematical proofs, and that this could jeopardize the peer-review system's ability to guarantee accuracy, transparency, and independent verifiability.

◆Second threat: Attribution of contributions and copyright issues
The declaration points out that while AI models learn by extensively utilizing publicly available mathematical papers and libraries, they may not properly cite human research findings in their output, which undermines traditional mechanisms of evaluation and accountability in mathematics.

◆The third threat: Dependence on AI technology and inequality
If researchers with access to the latest proprietary AI and expensive computing resources have an advantage, the recruitment, funding, and evaluation systems may become biased towards AI use itself, potentially putting researchers who cannot use or do not want to use such technologies at a disadvantage.

◆Fourth threat: AI-generated mathematical results are heavily promoted in press releases and blog posts rather than in peer-reviewed papers.
The declaration states that such announcements not only lead to an overestimation of AI's capabilities and an underestimation of prior human research and contributions, but also to the use of specific mathematical problems as an indicator of the general reasoning ability of commercial AI products.

◆The fifth threat: Loss of autonomy in mathematical research
The declaration states that as technology companies become more involved, there is a risk that problems that are easy to automate or that serve to highlight the company's products will take precedence over problems that have deep mathematical significance.

To address these issues, the declaration recommends mathematicians include a 'Disclosure of Tools and Computational Resources' section in their papers to explicitly state the use of large-scale language models, machine learning systems, proof assistance systems, mathematical software, and other resources. It also states that even when using AI, only human authors should be responsible for the correctness of proofs, the validity of arguments, and the completeness of citations to prior research.

The declaration further states that AI should not be treated as the author, and that credit and responsibility should remain with humans belonging to the mathematical community. Given the limitations of AI in providing proper attribution, it suggests that it is necessary to actively seek out and indicate the sources on which new results are based, and to clearly state when sufficient attribution is not possible.

Research institutions, academic societies, and non-profit research funding organizations are required to establish policies regarding AI-based paper submission and peer review, authorship, transparency, intellectual property, and codes of conduct. Regarding AI-generated results, it has been proposed that humans be able to explain the central arguments, that formal verification be required when necessary, and that theoretical and computational results be cross-referenced.

In addition, the declaration emphasizes that mathematical achievements should continue to be published in peer-reviewed journals, conference proceedings, and books, with press releases and blog posts playing only a supplementary role. Furthermore, it calls for support for low-resource technologies that are easily accessible to universities, national and international research institutions independent of industry, and individual researchers.

The declaration calls on governments and policymakers to strengthen legal protections for authors' rights, to avoid blindly accepting exaggerated claims by AI companies, and to incorporate the opinions of experts, including mathematicians, into policy decisions. It also calls for regulation of the AI industry and investment in public computing infrastructure, and that mathematical research should not become overly dependent on certain commercial technologies.

The Leiden Declaration has been formally endorsed by the International Mathematical Union (IMU). IMU Vice President Ulrike Tillmann stated that while AI opens up new and exciting opportunities, it also creates problems that cannot be ignored, and the future of mathematical research should be guided by human judgment, fair and transparent practices, and the shared values of the international mathematical community.