OpenAI accused of pressuring researchers as math feud escalates
Twenty-five mathematicians, all Fields Medal winners, the discipline's most prestigious honor, have signed an open letter warning that AI labs racing to be first to solve famous unsolved problems are putting proper attribution, verification and the human transmission of mathematical knowledge at risk. The letter argues that when a lab rushes out a solution to claim priority, there is often no time for a proper writeup, for isolating the new methods and ideas involved, or for citing the relevant work of others who came before.
The letter follows a specific dispute that broke into public view this week. NYU professor Tristan Buckmaster accused OpenAI of pressuring him not to credit a collaborator, who works at Anthropic, for solving an important math problem. Buckmaster also wondered whether OpenAI had fed the pair's work with its Codex coding tool into its own systems to produce a separate, groundbreaking proof over what he described as a marathon weekend of inference; that OpenAI proof remains unverified, according to the article.
On Thursday, OpenAI withdrew its sponsorship of a math event at CalTech after researchers at the university criticized the company, though the specific nature of their criticism is not detailed.
The signatories argue the stakes go beyond who gets credit for a proof. Without mathematicians willing to develop and integrate AI-generated ideas into the broader mathematical canon, they wrote, those ideas would never become fully alive, and the transmission chain between mathematicians would be lost. They also warn that frontier labs able to spend tens of millions of dollars on AI systems to beat human researchers to a discovery create an incentive for secrecy rather than open publication. Their letter follows the Leiden Declaration, published by a working group of mathematicians in June with recommendations for researchers, institutions and policymakers, and closes with a warning aimed past mathematics itself: the same pressures on attribution, verification and the purpose of research, they say, await every other field AI touches.
Key facts
- Twenty-five Fields Medal winning mathematicians signed an open letter warning that AI labs' race to solve famous problems threatens attribution and the open research culture in mathematics.
- NYU professor Tristan Buckmaster accused OpenAI of pressuring him not to credit an Anthropic affiliated collaborator on an important math problem, and questioned whether OpenAI used their Codex work to produce its own proof over a marathon weekend of inference.
- OpenAI withdrew its sponsorship of a CalTech math event on Thursday after criticism from researchers at the university.
- The letter follows June's Leiden Declaration, a document from a working group of mathematicians offering recommendations for researchers, institutions and policymakers.
- The signatories warn that frontier labs can spend tens of millions of dollars using AI to beat original researchers to a proof, an incentive toward secrecy rather than open publication.
Why it matters
The dispute is a live test of whether AI generated proofs can be integrated into a field that runs on credit, verification and slow transmission of ideas between people. If labs can win priority by spending heavily on inference rather than doing the underlying work, the letter argues, the incentive shifts from open publication toward secrecy, and unverified results start to count as breakthroughs before anyone has checked them.
Who it affects
The twenty-five Fields Medal winning mathematicians who signed the letter, NYU professor Tristan Buckmaster and his unnamed collaborator at Anthropic, OpenAI, and the CalTech researchers who criticized the company's sponsorship. The signatories argue the same dynamic will eventually reach every field where AI can attempt to substitute for human discovery, not mathematics alone.
How to use it
There is no product or price here. The practical takeaway for researchers and institutions is the one the Leiden Declaration and this new letter both push: agree on attribution and verification norms for AI assisted results before a dispute like Buckmaster's becomes routine, and read the Declaration's recommendations if you sit on the institutional or policy side of that question.
How solid is it
The account rests on TechCrunch's reporting and the open letter's own text, both treated by the article as established rather than disputed. Buckmaster's accusation against OpenAI is presented as his account; the article does not describe any response from OpenAI, and it calls OpenAI's own proof "unverified" without explaining what verification process or timeline would resolve that.
Risks and caveats
Several details the article omits: the identity of the twenty-five signatories, of Buckmaster's Anthropic affiliated collaborator, and of the CalTech researchers; the specific math problem Buckmaster and his collaborator solved; the content of CalTech's criticism; whether Buckmaster himself is one of the twenty-five signatories; and any dollar figure OpenAI itself has actually spent, as opposed to the general capacity the letter attributes to frontier labs. Treat the Buckmaster episode as one contested account until OpenAI's side of it surfaces.
“As in all creative professions, this raises severe attribution and plagiarism questions. Moreover, without the willing mathematicians who must take care of their development and integration into the mathematical canon, AI-conceived ideas would never become fully alive and the crucial human transmission chain between mathematicians would be lost.”
— the mathematicians' open letter