Mathematicians Issue Leiden Declaration as AI Claims 10 Major Open Problems
Updated
Updated · understandingai.org · Aug 4
Mathematicians Issue Leiden Declaration as AI Claims 10 Major Open Problems
3 articles · Updated · understandingai.org · Aug 4
Summary
A growing group of mathematicians is rallying around the Leiden Declaration, a framework drafted after a September 2025 conference to preserve research values as AI rapidly advances in the field.
Recent AI milestones sharpened that push: OpenAI said an internal Astra model solved 10 major open problems, after other systems reportedly cracked long-standing conjectures that had resisted humans for decades.
More than 20 mathematicians interviewed at the International Congress of Mathematicians split over the impact, with some saying AI already helps search literature, fill proof gaps and speed drafts, while others fear replacement and funding cuts.
Terence Tao and other researchers argue the bigger risk is not just lost jobs but mathematics becoming less humanly understandable if AI proves results no one can explain, weakening training and the shared expert community.
The declaration is still an opening step rather than a full blueprint, reflecting a broader view that mathematics will survive AI disruption only if the field redefines its goals beyond simply solving problems.
If AI autonomously conquers open Erdős problems, is the traditional mathematics PhD now effectively obsolete?
Why are top Fields Medalists abandoning graduate mentorship as AI begins to rewrite the rules of mathematical discovery?
When machines prove theorems no human can comprehend, does mathematics lose its fundamental purpose of human understanding?
OpenAI Astra Solves Ten Landmark Math Problems: The Verification Crisis and the Battle for Human-Centric Mathematics
Overview
In 2026, AI models like OpenAI's Astra began solving major mathematical problems, triggering a dramatic shift in research and causing top mathematicians to move from academia to industry. This rapid progress led to anxiety among students and early-career researchers, as AI quickly outpaced human abilities and made traditional education and career paths uncertain. To address concerns about integrity and control, the mathematical community released the Leiden Declaration, warning against overreliance on closed, corporate AI systems. Meanwhile, OpenAI formalized Astra's proofs using Lean and published machine-checkable certificates, collapsing peer review timelines but raising new questions about human understanding and the future of open science.