Terence Tao-Led Mathematicians' Group Urges Boycott of OpenAI Over Proof Release

The dispute escalated after OpenAI, a company facing lawsuits over illegal copying, copyright infringement and trademark dilution, released manuscripts claiming solutions to well-known mathematical problems. The AHM statement said no one had asked OpenAI to solve these proofs. Earlier, after OpenAI said it had solved a Millennium Prize problem, Tao and 25 Fields Medalists issued a joint appeal asking commercial companies to slow down, warning that AI companies were accelerating without knowing what would follow. OpenAI did not heed the appeal.

After the Millennium Prize plagiarism controversy, OpenAI worked with the Princeton Institute for Advanced Study to establish the Mathematics and Artificial Intelligence Advisory Group, AGMAI, and said it would build deep cooperation with the mathematics community. According to QuantumBit, the AHM statement described that cooperation as a public relations gesture and said the expert group's views were ignored. The group's first advisory report warned that frontier AI companies should not test advanced mathematical problems on internal models. OpenAI ignored the warning, used an internal model to test about 8,000 open mathematical problems, achieved a hit rate of about 5 percent and released 700 machine-generated drafts. The AHM statement said this disregard showed contempt for scientific research norms and rejected OpenAI's claim that the release advanced mathematics.

Theoretical computer scientist Scott Aaronson called the release a Mathocalypse. He mentioned his wife, Dana Moshkovitz, who has devoted her academic career to the Unique Games Conjecture, a major problem in theoretical computer science. One of the documents released by OpenAI stated that the conjecture had been proved. After reading the documents overnight, Moshkovitz said she felt absurdity and despair. In a message, she wrote that it felt like something written by someone on hallucinogens. The paper, she said, was full of chaotic logical leaps, piled up prior conclusions without explaining their applicability and used strange constructions never seen in human mathematics. It was so poorly written that it could not be understood without AI assistance, she said, and contained alien-like logic. The article said there was no peer review, no readable human logic and no respect for previous work.

QuantumBit also said the released output notably lacked cryptography, and that there were indications AI companies had been secretly using powerful models to probe underlying vulnerabilities in modern cryptographic protocols and primitives.

Tao had previously warned of proof indigestion. He called the destruction irreversible pollution: once a problem is claimed to be solved, it can never return to an unsolved state. Even knowing that an answer exists, he said, can pollute other research paths that might lead to deeper insight. Many young scholars, fearing that large models will scoop them, do not dare to publicly reveal the directions they are exploring. In 2024, Tao imagined industrialized mathematics optimistically and believed humans would remain the leaders, with AI serving as an assistant like a chess engine. In a short time, the balance has reversed: AI solves problems beyond human reach, while humans become tools for checking calculations and organizing notes.

Turing Award winner Yann LeCun commented on the conflict. His attitude appeared different from that of Tao and other prominent mathematicians. He said: 'On the contrary. Mathematics is entering a new era. Formal proofs will be largely automated, and the focus will shift to developing new concepts, new abstractions, new definitions and new conjectures. The invention of ships reduced the importance of swimming, but let us discover new continents.'