OpenAI has faced criticism from the mathematical community for announcing AI-solved problems via social media and blog posts rather than peer-reviewed scientific journals. Researchers argue this practice hinders verification and obscures the contributions of human scientists. In response, OpenAI released 722 mathematical manuscripts on GitHub, categorized into 372 families of results, which were generated by an internal model that evaluated approximately 4,000 problems.
The company claims its system has solved over 100 long-standing mathematical problems, including the quasi-Riemann hypothesis and the rational Hodge conjecture. These results required an average of three hours of computation per task using ChatGPT Pro. OpenAI has established an Advisory Group on Mathematics and Artificial Intelligence (AGMAI) to guide its disclosure processes, though experts note that the group's September recommendations—such as using independent repositories—have not yet been fully implemented.
Tensions have escalated following allegations that OpenAI used thousands of AI agents to preempt independent research on millennium problems. Mathematicians have expressed concerns over the company's competitive tactics, with some researchers reporting pressure regarding authorship and career implications. While OpenAI maintains that its technology aims to expand the capabilities of mathematicians, the scientific community continues to call for more transparent and collaborative publication standards.
