All the drama around AI’s takeover of mathematics

This year, AI research labs such as OpenAI and Anthropic have publicized a string of major advances in mathematics, claiming solutions to problems that many considered out of reach for current systems — including one of the Millennium Prize problems. Rather than being greeted solely with celebration, these announcements have provoked controversy among professional mathematicians over how the work was produced and disclosed. A recurring criticism centers on transparency: researchers have pressed OpenAI to clarify what training material and unpublished work its models used. In one exchange, Andreas Thom publicly questioned whether prior interactions with ChatGPT contributed to the company’s recent successes. OpenAI later acknowledged that one reported result built substantially on earlier work by Thom and Gábor Kun, intensifying calls for clearer attribution and data provenance. OpenAI has attempted to repair relations by creating a small independent advisory panel of mathematicians charged with guiding how it and other companies engage with the field and release findings. Several mathematicians told reporters the effort was a step forward but left many practical questions unanswered, including what authority the panel will have and whether a handful of prominent members can represent the broader research community. Beyond process and transparency, critics say the labs’ competitive rush to publish may be changing norms of the discipline. Some mathematicians describe tactics such as last-minute pushes to announce results as breaching established academic etiquette, and there have been allegations of scooping and other missteps. AI companies say they are learning from earlier errors, but the mathematical community remains watchful about whether those promises will translate into more collaborative and transparent practices going forward.