OpenAI publishes 722 AI-generated math manuscripts on GitHub

OpenAI has published 722 AI-generated math papers that were created using an unreleased artificial intelligence model. SiliconANGLE reported that the company posted the documents to GitHub late on Tuesday. As FTMQ SI reported earlier, the manuscripts contain solutions to open mathematical questions generated by an unreleased frontier system. [1][2][3][4][13]

Some of the published papers prove long-running hypotheses, according to SiliconANGLE. Other manuscripts disprove proposed explanations of mathematical phenomena or narrow down potential answers to unsolved problems. SiliconANGLE noted that the papers span about 20 mathematical subfields, including a discovery related to the Riemann hypothesis, a paper over 150 years old concerning prime numbers. [1]

Reports from Yellow.com and Gadget Review stated that independent mathematicians and experts must now verify and judge the published work. Quartz noted that the release followed a claim regarding the Millennium Prize. Earlier FTMQ SI reporting showed that the 722 manuscripts cover 372 result families and have raised concerns among academic researchers over research ethics and attribution. [5][7][8][13]

OpenAI is an American artificial intelligence public benefit corporation headquartered in San Francisco, California. The company previously released models including GPT-5 in August 2025, followed by GPT-6 Astra and GPT-6.1 Sol in 2026. Academic and commercial researchers are reviewing the latest mathematical release to evaluate its scientific validity. [10][12][13]