25 Sep 2026

ProNewsFlash

AI policy, regulation and law

Tao Warns AI May Spark Math’s Biggest Crisis Since Gödel

Mathematician Terence Tao argues in a new essay that AI-generated proofs could force the field to confront long-unspoken questions about what mathematical research is actually for.
Tao Warns AI May Spark Math’s Biggest Crisis Since Gödel

Mathematician Terence Tao warns that artificial intelligence could push mathematics into a crisis on par with the foundational upheaval of the early 20th century, according to an essay reported by the-decoder.com. Tao wrote the piece for the 2026 International Congress of Mathematicians.

The-decoder.com reports that Tao draws a parallel to the period between 1900 and 1930, when Russell’s paradox and Gödel’s incompleteness theorems forced mathematicians to make their implicit assumptions explicit, eventually producing a rigorous framework that has lasted a century. Today, Tao argues, the stress test is not about mathematical truth but about implicit values and practices, including what counts as a contribution and whether a machine can be said to have done the work.

Tao’s working hypothesis is that AI tools will soon handle a meaningful share of research-level mathematical tasks. He points to the First-Proof Project, in which seven of ten unpublished research problems received at least one passing grade from AI systems tested under controlled conditions, at costs of tens to hundreds of dollars per problem, per the outlet.

According to the-decoder.com, Tao warns the field could shift from proof scarcity to proof abundance, with AI-generated proofs accumulating faster than experts can verify them. He also cautions that overly polished AI proofs can strip away the natural friction found in human-written work, making them harder to learn from.

Tao points to the Leiden Declaration on Artificial Intelligence and Mathematics, published in June 2026 and backed by the International Mathematical Union, for guidance. His own standard: if authors cannot give a clear, correct, properly attributed expert-level talk on their results, the result should not be published, the outlet reported. Tao also stresses protecting the human element of training young mathematicians and discloses his own limited AI use for tasks like literature search and formatting.

Based on reporting by the-decoder.com.