OpenAI releases over 700 AI-generated mathematical proofs, retracts three papers

Oct 7, 2026

On October 6, 2026, OpenAI released over 700 AI-generated mathematical proofs, prompting intense debate and calls for a boycott from the Association for Mathematical Heritage. While some experts like Alex Kontorovich and Martin Bridson call the breakthroughs breathtaking and worthy of a Fields Medal, others, including Fields Medalist Terence Tao and researcher Dana Moshkovitz, warn of a devastating impact on the discipline. They argue that mass-produced, poorly written AI proofs force human mathematicians into the thankless role of mere verifiers, threatening the traditional pursuit of conceptual understanding.

Details of OpenAI's release and retractions

▪OpenAI tested approximately 8,000 mathematical problems using an internal model at the GPT-Pro level, achieving a success rate of roughly five percent with an average of three hours of compute per problem
▪OpenAI released over 700 AI-generated mathematical proofs in a single batch on October 6, 2026, which included solutions to open problems across algebra, number theory, theoretical computer science, mathematical logic, and topology
▪OpenAI retracted three of its released mathematical papers shortly after publication

Boycott of OpenAI and subsequent debate

▪Mathematicians on Terence Tao's blog debated the Association for Mathematical Heritage's boycott of OpenAI, with some supporting it and calling for preprint servers like arXiv to block training data collection, while others called the boycott unrealistic
▪The Association for Mathematical Heritage called for a boycott of OpenAI, describing the October 6, 2026 release of over 700 files as a demonstration of power rather than scholarship

Warnings on the misalignment of AI and mathematics

▪Terence Tao and 24 other Fields Medalists signed a statement warning of a severe misalignment between the AI industry's goals and those of mathematics, criticizing rushed announcements that raise plagiarism questions
▪The Advisory Group on Mathematics and AI (AGMAI), established by OpenAI at the Institute for Advanced Study, took a diplomatic stance, warning that the future of math cannot consist solely of verifying AI-generated results

Navier-stokes controversy

▪OpenAI claimed that an internal model solved more than 100 open math problems in a single month, including an approach to the Navier-Stokes Millennium Problem
▪OpenAI denied suspicions that it used data from ChatGPT interactions with two mathematicians to scoop their AI-assisted solution to the Navier-Stokes Millennium Problem

Debatable claims

▪OpenAI's batch-release of raw proofs does more harm than good
▪Mathematicians should boycott OpenAI over its mass release of proofs
▪Verifying AI-generated results is an inappropriate role for human mathematicians
▪Rushed AI proof announcements pose an unacceptable risk of plagiarism