Back to News
NewsJobsAi

OpenAI Releases Math Papers from Unreleased AI Model, Faces Skepticism

OpenAI's release of 722 math papers raises skepticism and debate over AI's role in mathematical research and the need for human verification.

Mercation Admin
October 7, 2026 2 min read
OpenAI Releases Math Papers from Unreleased AI Model, Faces Skepticism

OpenAI Unveils 722 Mathematics Papers, Faces Mixed Reactions

OpenAI has made headlines by releasing 722 mathematics papers on GitHub, generated by an unreleased internal AI model. While the announcement suggests a significant leap in mathematical research, the response within the academic community has been cautious.

According to OpenAI, most of the results were produced from a single prompt to the AI, although some required multiple attempts. However, MIT mathematician Andrew Sutherland warns that without public access to the model, claims of solving problems with a single prompt remain unverified. “Until they provide access to the model, any assertions should be considered unverified,” he stated.

The collection is organized into 372 families of results, where the total of 722 refers to manuscripts rather than distinct solved problems. OpenAI posed around 4,000 problems to the model, with the average result taking about three hours of computational time using ChatGPT Pro.

Out of the 722 papers, only 162 contain a verified main result, which is roughly 22% of the total. These results were checked using Lean, a software that confirms logical steps but does not ensure the originality or significance of the findings. OpenAI cautions that many unverified results may contain errors, raising concerns about the reliability of the published work.

Researcher Dmitry Rybin expressed skepticism about the clarity of some proofs, describing them as "incomprehensible alien math." Meanwhile, Keith Adler criticized the lack of engagement on the GitHub repository, noting that it has not accepted community contributions despite the volume of manuscripts released.

Concerns have also been voiced by the Institute for Advanced Study, which emphasized the need for human oversight in mathematical reasoning. The institute worries that AI-generated arguments might be beyond comprehension or verification by humans, complicating academic accountability.

Despite the skepticism, some mathematicians, like Professor Abhishek Saha, view the release as a significant milestone, albeit with most findings categorized as advancements within existing frameworks rather than groundbreaking discoveries. Only the Quasi-Riemann Hypothesis is seen as a truly notable result among the 722 papers.

The release did not fully align with recommendations from an advisory group at the Institute for Advanced Study, which had called for detailed documentation of the model and its findings. While OpenAI has promised to provide more Lean formalizations in the future, the current lack of transparency has led to calls for better practices in disseminating AI-generated research.