Mathematicians call for OpenAI boycott after AI research release covers 377 maths problems

OpenAI is facing a boycott call from the Association for Human Mathematics after publishing findings on 377 mathematical problems and releasing a repository containing more than 700 files. The group says mathematicians did not request the work and has accused the company of disregarding scientific research norms.

The dispute comes amid growing debate over the role of artificial intelligence in advanced mathematics. It follows OpenAI’s claim last month that one of its internal models had solved the Navier-Stokes Millennium Prize problem, a famous challenge in mathematics.

Why are mathematicians calling for an OpenAI boycott?

In a statement responding to OpenAI’s October 6 release, the Association for Human Mathematics challenged the company’s approach to AI-generated research.

“Mathematicians did not ask for this work to be done,” the group said, questioning OpenAI’s claim that its findings would advance mathematics. It described the release of hundreds of files as “not a demonstration of scholarship, but a demonstration of power”.

The association also argued that OpenAI had ignored concerns raised by the Advisory Group on Mathematics and Artificial Intelligence, which has cautioned against frontier AI companies testing advanced mathematical problems on internal models.

The group urged mathematicians and the public to remain sceptical of the company’s approach.

What are the concerns about AI in mathematics?

The Advisory Group on Mathematics and Artificial Intelligence, hosted by the Institute for Advanced Study in Princeton, raised concerns in a September 29 statement about AI-generated mathematical arguments. It warned that AI systems could produce reasoning that human researchers might struggle to understand, verify or take responsibility for. Another concern is unequal access. If AI companies have access to powerful models and computing resources unavailable to independent researchers, mathematics could develop a “two-tier system”, giving some researchers a significant advantage. The debate is therefore not just about whether AI can solve difficult problems. It is also about how results are produced, checked and shared, and whether researchers have a fair opportunity to participate.

How has OpenAI defended its mathematical research?

OpenAI says it has consulted the advisory group and considered its advice and public recommendations when deciding how to release its latest findings.

The company published its results on GitHub, including formal versions of many proofs written in Lean, a programming language that allows mathematical proofs to be checked by a computer. OpenAI also shared 10 summaries of its model’s reasoning, information about the number of problems attempted and estimates of the computing resources used. It said the average result required the equivalent of around three hours of ChatGPT Pro thinking. These materials provide ways to examine the work, but the wider disagreement over research standards, transparency and access remains.

The controversy highlights a central question for AI research: how can mathematical advances be assessed and trusted when powerful systems are developed and tested within private companies?