A group of 25 prominent scholars has endorsed an open statement warning that artificial intelligence companies are undermining their academic discipline as they race to outdo one another by cracking celebrated mathematical mysteries. Every signatory of the document has received the Fields Medal, widely regarded as the highest honor in the field of mathematics.
Earlier this week, NYU professor Tristan Buckmaster alleged that OpenAI urged him to omit recognition for a co-researcher employed by rival firm Anthropic after cracking a major problem. He also questioned whether the AI giant leveraged their shared activity on Codex to generate its own milestone proof during an intensive compute-heavy weekend.
By Thursday, OpenAI had pulled its financial support for a mathematics conference at CalTech following pushback from academics at the institution.
Although AI systems capable of resolving humanity’s most complex mathematical puzzles could offer immense benefits, the authors of the letter stress that such breakthroughs are only useful if researchers can actually comprehend, verify, and share those insights with the broader public.
“Frequently, these findings are unveiled in haste, skipping essential steps like detailed documentation, identifying novel techniques, and acknowledging prior research,” the researchers highlighted, noting that OpenAI’s recent proof has yet to be formally validated. “Just as in any creative discipline, this triggers serious concerns regarding academic credit and plagiarism. Furthermore, without dedicated mathematicians to unpack, refine, and incorporate these concepts into established theory, machine-generated discoveries cannot truly mature, breaking the vital chain of human knowledge transfer.”
As anxieties mount over whether academic work performed via Codex is secretly training OpenAI’s next-generation architectures, experts worry that the legacy of open collaboration is under siege. Modern frontier labs possess the resources to sink tens of millions of dollars into large language models to scoop original discoverers—a trend that inevitably pushes researchers toward greater secrecy.
The statement comes on the heels of the Leiden Declaration published by a panel of mathematicians this past June. That initiative similarly addressed how machine-generated proofs are reshaping academia, proposing actionable guidance for scholars, academic organizations, and legislators.
Much like in programming and other domains transformed by automation, mathematicians argue that the true essence of their field extends beyond final answers: the core value of mathematics lies not simply in claiming credit for proofs, but in sustaining the intellectual ecosystem that trains future thinkers, poses groundbreaking questions, and weaves knowledge into society.
Even if high-stakes mathematical disputes seem distant, take note: similar disruptions are bound to reach your own industry soon.
“The challenges currently confronting mathematicians mirror those in other scientific and artistic domains, offering a preview of what society at large may soon encounter: ensuring that as artificial intelligence alters how work is performed, we preserve the fundamental purpose behind that work,” they concluded.
