Home 9 AI 9 OpenAI’s Math Claims Put Research Trust to the Test

OpenAI’s Math Claims Put Research Trust to the Test

by | Sep 29, 2026

An independent advisory group faces questions about its role as mathematicians await a wave of AI-generated results.
Source: The Verge, Shutterstock.

 

OpenAI’s advances in mathematics have created a problem that proofs cannot solve: researchers need time, context, and credit to assess new results. According to The Verge (full article available to subscribers), the company hopes to improve its relationship with mathematicians by consulting the Advisory Group on Mathematics and Artificial Intelligence, or AGMAI. Yet the group’s debut has already exposed confusion about its role.

Nine mathematicians formed AGMAI after OpenAI approached some of them about an advisory board. The group says it operates independently and can advise other AI laboratories. Member Martin Hairer told The Verge that it receives no funding or technical support from OpenAI. Its members may criticize the company publicly, although they will not set the pace of its internal research. OpenAI’s announcement nevertheless led some mathematicians to mistake AGMAI for a company-appointed panel.

The stakes are growing. OpenAI says an unreleased model has resolved more than 100 longstanding mathematical problems. Researchers interviewed by The Verge worry about a sudden flood of claims that would demand effort to verify and understand. Some also fear that undisclosed results could overtake work they have pursued for years.

Past releases help explain the mistrust. Mathematicians criticized manuscripts they considered poorly written, thin on relevant scholarship, or unclear about earlier contributions. The Verge also reported instances in which OpenAI revised documents without recording the changes. Such practices make it harder to judge findings and assign credit.

AGMAI could help AI companies communicate results responsibly, but its independence and influence will be judged through its work. The article emphasizes that a mathematical answer is only part of a discovery. Researchers must establish why it matters, connect it to knowledge, and give others a transparent account they can check. Managing those obligations will determine whether AI-generated proofs strengthen mathematics or leave scholars sorting through claims at an unmanageable pace.