OpenAI AI Model Reshapes Mathematics Research Debate

OpenAI has released hundreds of AI-generated mathematical manuscripts that claim solutions to difficult problems, including questions studied for decades, prompting leading mathematicians worldwide to debate the impact of artificial intelligence on research. The results were published on October 6, 2026, through a private internal model that generated proofs across multiple fields of mathematics.

According to reports from Le Devoir, Le Figaro, L'Express, and Les Numeriques, OpenAI submitted about 4,000 problems to its model and published 719 manuscripts covering 372 families of results. The company said the system produced solutions in hours, while many human researchers had spent years working on similar questions.

OpenAI's results include problems related to algebra, number theory, topology, probability, and statistical physics. The company said about 42 percent of the main results had formal verification through Lean, a software system that checks mathematical proofs line by line, while other results still require human review.

Hugo Duminil-Copin, a 2022 Fields Medal winner, professor at IHES and the University of Geneva, said the announcement affected his research area because 16 problems connected to his field appeared among the AI-generated results.

"On Tuesday night, I cried and it woke up my daughter," said Hugo Duminil-Copin, according to reports from Le Figaro and Les Numeriques.

The mathematician said he plans to examine more than 3,500 pages of material generated by the AI system. He also questioned the personal connection researchers feel toward proofs created by machines rather than colleagues.

"I do not really want to read these proofs, because doing so does not have the same flavor as if they had been written by humans whose faces I know," said Hugo Duminil-Copin, according to Le Figaro.

Luc Vinet, a professor of physics at the University of Montreal who has studied mathematics since the 1970s, described the development as a major shift for the field. He said AI could also create new opportunities if researchers maintain focus on understanding results.

"Mathematical research will never be the same again," said Luc Vinet, according to Le Devoir.

Vinet said he has used AI in his own research and recently applied it to the final step of a problem he had worked on for more than 20 years. He has not yet published that solution because he wants to fully understand the result.

"It can lead to a renaissance of mathematics," said Luc Vinet, according to Le Devoir.

He emphasized that researchers must interpret and explain AI-generated discoveries before they become part of mathematical knowledge.

"I now have to work very hard to reclaim this result, to humanize it," said Luc Vinet, according to Le Devoir.

Marcel Goh, a doctoral candidate in mathematics at McGill University, said he avoids using AI tools in his research because he believes the process of proving theorems is central to learning.

"Proving theorems is a game that we play to improve ourselves. The goal is never to win the game," said Marcel Goh, according to Le Devoir.

Goh is among members of the Association for Human Mathematics, a group founded by mathematicians that has more than 800 members. The organization criticized OpenAI's release process and called for greater emphasis on human understanding.

"The simultaneous publication of more than 700 files is not a scientific demonstration, but a demonstration of power. We urge mathematicians to end their collaboration with OpenAI and return to a vision of science that places human understanding at the center of its concerns," said the Association for Human Mathematics in a statement cited by Le Devoir.

Cedric Villani, a Fields Medal winner, described the development as a historic change in mathematics. He said the results could represent a scale of discovery comparable to major advances recognized in the field.

"This is certainly the most radical change in mathematics for at least four centuries, perhaps even 1,000 years," said Cédric Villani, according to Le Figaro.