Mathematicians Publish Leiden Declaration on AI Risks to Mathematics

On 2 June 2026 the Leiden Declaration on Artificial Intelligence and Mathematics was published, per the declaration website. The document, developed after a 2025 workshop at the Lorentz Center in Leiden, sets out values and recommendations for the mathematical community and related institutions, the Leiden Declaration text states. Gizmodo reported the declaration had attracted more than 130 signatories at the time of publication. The London Mathematical Society and Scientific American summarised the declaration's core concerns as including the reliability of automatically generated results, attribution of work involving proprietary models, and effects on publication and peer review.

The Leiden Declaration frames its concerns around the role of proof as central to mathematical practice and the norms of attribution and responsibility for correctness, per the declaration preamble. The text discusses both symbolic and neural methods for generating and formalising mathematics, and highlights tensions between rapid public dissemination of AI-generated claims and established procedures for validation, according to the declaration and reporting in Gizmodo and Scientific American. The document recommends that authors disclose AI assistance, ensure results are peer reviewed, and that funders and professional bodies develop policies to preserve scrutiny and accessibility, as described on the LMS summary and the declaration site.

The issues the declaration raises mirror broader debates in research communities about reproducibility, provenance, and commercial control over model weights and training data. Observed patterns in similar domains show that when automated tools accelerate claim generation, peer-review bottlenecks and incentives for rapid publicity often increase. For practitioners, this tends to shift effort toward verification, provenance tracking, and investing in tools for reproducible evaluation rather than purely novel discovery.

Editorial analysis: For the mathematics community the declaration is notable because it treats mathematical proof and attribution as institutional norms worth defending, framing AI not simply as a tool but as something that can reshape practices of credit and validation. Media coverage (Scientific American, Gizmodo) links the declaration to a recent rise in AI-generated proofs submitted to journals and preprint servers, which editors and some mathematicians report are increasing editorial workload. The declaration also engages policy actors: it asks funders and policymakers to consider public investment and regulatory attention, a framing that aligns the document with other cross-disciplinary calls for responsible AI governance.

Editorial analysis: Observers will track whether major mathematical journals and conference series adopt explicit AI-disclosure and verification policies, and whether funders require reproducibility and open access to artefacts. Reporting by the LMS and the declaration itself suggests three near-term indicators to follow:

• •adoption of disclosure requirements by high-profile journals;
• •development of infrastructure for independent verification of formalised proofs; and
• •policy statements or guidance from national research funders on use of proprietary models in published research.

Editorial analysis: For AI practitioners building math-capable systems, the declaration signals increased demand for tooling that produces verifiable, attributable outputs, for example, systems that produce machine-checkable proof objects or provenance metadata. It also indicates potential political and funding pressure toward open, auditable models and datasets in research contexts.

Reported reaction has been mixed. Gizmodo quoted Christoph Sorger, secretary general of the International Mathematical Union, saying, "I do not expect every colleague to agree with every sentence of the declaration," in IMU's endorsement of the text. Commentary from community blogs (for example, Quantum Formalism) framed the declaration as a necessary defence of standards rather than anti-AI rhetoric. The Newton Institute coverage emphasises the document does not call for a blanket ban on AI, describing it instead as a call for collective action to ensure technology supports the discipline.