OpenAI releases 722 AI-generated math manuscripts, sparking ethics debate

Massive batch of AI-generated proofs

OpenAI announced a GitHub repository containing 722 manuscripts that it says solve “hundreds” of open questions, grouped into 372 result families The Verge. The company previously claimed more than 100 long-standing problems were resolved, and the new release adds summaries of model reasoning, compute estimates and problem-attempt statistics. OpenAI notes the “average result” required roughly three hours of ChatGPT Pro-level processing.

Advisory group’s response

The newly formed Advisory Group on Mathematics and Artificial Intelligence (AGMAI) published recommendations in late September, urging AI labs to release results through established academic channels, disclose model names, prompts and compute costs, and avoid using breakthroughs as marketing tools. AGMAI argues that treating releases as promotional material “inflicts significant harm on the mathematical community.”

Ethical and credit-allocation concerns

The batch includes work touching a Millennium Prize problem, one of the field’s most celebrated open questions. The rapid succession of AI-driven announcements this year has sparked fierce debate over research practices, the ethics of publishing AI-generated proofs, and how companies credit the human mathematicians whose prior work may have informed the models. Rival labs such as Anthropic have produced comparable outputs, expanding the pool of AI-originated mathematics that scholars must evaluate.

Our read: OpenAI’s unprecedented scale of AI-generated mathematics forces the community to confront how proof validity, peer review and attribution will be managed when machines produce the bulk of new results.