Story
October 10, 2026

OpenAI’s Math Deluge Dazzles Researchers—and Raises a Fight Over Control

OpenAI’s release has exposed a split in mathematics: enthusiasts see a powerful new engine for discovery, while critics and advisers warn that speed, scale and corporate control could erode the human judgment that gives the discipline its meaning.

The fault line opened in September, when OpenAI said an internal model had resolved more than 100 longstanding problems, including an apparent solution to Navier-Stokes. The claim stirred excitement, but also accusations that the company’s approach risked outrunning academic scrutiny.

In response, the Institute for Advanced Study-hosted Advisory Group on Mathematics and Artificial Intelligence urged labs to publish promptly through established channels, disclose models, prompts and compute, and avoid turning mathematical releases into marketing campaigns. Its warning was blunt: such promotion could cause “significant harm” to the mathematical community.

Then, on Tuesday, OpenAI released 722 manuscripts spanning 372 families of results—hundreds of open questions across fields from algebra and topology to number theory. The scale stunned researchers. Oxford mathematician Martin Bridson called the release “breathtaking,” saying it had become possible for mathematical frontiers to move dramatically “in one day.” OpenAI co-founder Greg Brockman framed the effort as moving “towards acceleration of scientific discovery and improving quality of life for everyone.”

Supporters see an opening rather than an obituary for mathematicians. University of Toronto professor Dan Litt called the new results “great for mathematics,” while warning that society must keep supporting human expertise if it hopes to benefit from AI’s progress. Yann LeCun likewise described a new era in which formal proof is largely automated and mathematicians can focus on concepts, abstractions and definitions.

But the backlash intensified as scholars confronted the workload. More than three dozen mathematicians told The Verge that making sense of the material could take years, even before judging its originality or correctness. The Association for Human Mathematics called the mass release “not a demonstration of scholarship, but a demonstration of power.”

By Thursday, OpenAI had withdrawn results for at least three problems after apparent errors. The episode sharpened the central dispute: AI may generate more mathematics than ever, but humans still have to decide what is true, important—and worth building on.

Story coverage