An international group of mathematicians has issued the Leiden Declaration warning that artificial intelligence threatens the integrity of mathematical proofs [1].

The movement signals a growing concern that the speed and automation of AI could erode the rigor and trust essential to scientific discovery. If the core values of the discipline are compromised, the foundation of mathematical truth may be undermined.

The declaration is backed by a coalition of experts, including more than 150 professors [1] and more than 130 top mathematicians [2]. These signatories said that the current trajectory of AI integration risks compromising the attribution of work and the autonomy of research [3].

Central to the concern is the potential for AI to produce results that lack the transparency required for traditional verification. The group said that AI could threaten the very foundations of mathematics by prioritizing output over the logical process of proof [2]. This shift could lead to a crisis of trust in scientific literature, where the origin of a discovery becomes obscured by machine generation [3].

The Leiden Declaration emphasizes the need for strict safeguards to maintain human oversight. The signatories said that the discipline must protect the rigor of proofs and the intellectual autonomy of the researchers involved [3]. Without these protections, the group said that the pursuit of mathematical truth could be replaced by a reliance on black-box algorithms [2].

This collective action comes as AI tools become more prevalent in academic settings. The mathematicians involved said that the hype surrounding AI capabilities often masks the risks to the integrity of the field [1]. They said that a framework is needed to ensure AI remains a tool for human mathematicians rather than a replacement for the critical thinking required to validate complex theorems [3].

“AI could threaten the foundations of mathematics”

The Leiden Declaration represents a formal pushback against the 'black box' nature of generative AI in high-stakes academic fields. By focusing on attribution and proof integrity, the mathematical community is attempting to establish a boundary between computational assistance and the cognitive rigor required for scientific validation, ensuring that human logic remains the final arbiter of truth.