OpenAI math claims called power display, not scholarship

OpenAI's release of 722 AI-generated "mathematical results" in October 2026 triggered an unusually direct rebuke from the Association for Human Mathematics, a body made up of mathematicians worldwide. Central to the dispute is how OpenAI framed the release: the company said it consulted AGMAI, the Advisory Group on Mathematics and Artificial Intelligence hosted by the Institute for Advanced Study, and built the publication on AGMAI's September 29, 2026 proposal on responsible release of AI-generated mathematical information.

The association's objection turns on the text of that same proposal, which opens by saying that while some frontier AI research organizations verify advanced mathematical problems using proprietary models unavailable to the wider scientific community, such conduct should ideally be avoided and should be halted. In the association's reading, OpenAI used AGMAI's name as a source of legitimacy while disregarding the guidance itself. The association also called on mathematicians to stop joint research with OpenAI.

Reporting around the release adds further friction. TechCrunch found that only 10 of the 719 results disclosed the AI model's reasoning, with 3 later withdrawn, and a paper by mathematicians including researchers at the University of Cambridge flagged problems in the formalization of proofs using Lean. A separate thread of coverage warns that AI models solving advanced math could threaten the encryption that keeps internet traffic private, with Ethereum researcher Justin Drake saying the announcement shows mathematical superintelligence is becoming real.