OpenAI Creates Math Advisory Group After AI Model Solves Over 100 Open Problems

OpenAI has formed an advisory group focused on mathematics, following internal claims that its AI systems have resolved more than 100 previously open problems in the field. The company confirmed the move this week, framing the panel as a way to bring outside mathematical expertise into its research process. The group's creation signals how quickly AI-assisted proof work has moved from curiosity to something OpenAI wants formal oversight for.

The scope of that oversight is narrow by design. According to OpenAI, the advisory group will not have the authority to halt or redirect the company's ongoing mathematical research. Advisers can offer guidance and review, but the research roadmap stays under OpenAI's control. That arrangement puts the panel closer to a consulting body than a governing one, a distinction likely to draw scrutiny from mathematicians who have pushed for independent review of AI-generated results.

The 100-plus figure is the headline number, and it is also the one most likely to face outside verification. Mathematicians have spent decades building a culture in which a claimed proof means little until peers can check every step. AI systems can produce plausible-looking arguments that collapse under close reading, and OpenAI has not released a full accounting of which problems were solved, how difficult they were, or whether the results have been submitted for peer review. Until that record exists, the tally remains a company claim rather than a settled fact.

The advisory group fits a pattern across major AI labs, which have increasingly assembled ethics, safety, and domain-expert panels as their models reach into technical territory once reserved for trained specialists. Mathematics is an unusually sensitive case because progress there is judged by proof, not by benchmark scores. A result that is 99 percent correct is still wrong, and the field's norms make it hard to wave away errors as acceptable margins.

What happens next depends on transparency. If OpenAI publishes the problems its systems tackled, names the advisers, and submits results to journals or proof assistants for checking, the panel could become a genuine accountability mechanism. If it stays advisory in name while the company keeps its findings private, the group may look more like reputational cover than oversight. The open question is whether mathematicians outside OpenAI will treat the panel's work as credible evidence, or wait for proofs they can verify themselves.