OpenAI Pulls Sponsorship of Caltech Math Event After Letter From 25 Fields Medalists

The dispute between OpenAI and parts of the mathematics community escalated this week. On Thursday, September 10, 2026, the company announced it was withdrawing its sponsorship of the Mathathon competition organized by Caltech, and a day later 25 mathematicians, each a Fields Medal laureate, the most prestigious honor in the field, signed a separate open letter.

Mathathon was meant to be a competition in which teams would spend 40 hours working on open research problems with the help of large language models, using tokens funded by sponsors. The first round was scheduled for October 30 on the Caltech campus. OpenAI and Anthropic, as sponsors, were to provide a combined $2 million in computing credits, with OpenAI covering half of the budget allocated to each team.

Two letters, two accusations

Criticism of the event came from two directions. The first letter, published on proofsandprompts.com by current and former mathematicians affiliated with Caltech, accused AI companies of engaging in what it called a scientific disinformation campaign and warned of a phenomenon dubbed slop mathematics, AI-generated results that look credible but contribute nothing to the advancement of mathematical knowledge.

The second letter, hosted at mathandai.org, was signed by 25 mathematicians, each a recipient of the Fields Medal. The document does not directly criticize the Caltech event itself, but more broadly describes the tension between what AI models are optimized for, quickly producing true statements, and what mathematics as a discipline is meant to serve: developing conceptual understanding, insight, and training future generations of researchers.

Often these solutions are announced in a rush, leaving no time for a proper writeup - from the open letter signed by 25 Fields Medal laureates, mathandai.org

As in all creative professions, this raises severe attribution and plagiarism questions - from the open letter signed by 25 Fields Medal laureates, mathandai.org

OpenAI's response

OpenAI decided to withdraw as a Mathathon sponsor after criticism from researchers affiliated with Caltech. The decision was announced on X on Thursday, September 10, by Dan Roberts of OpenAI, who wrote that the company recognizes that AI's rapid progress in mathematics is disruptive to the field's established norms.

We recognize that the rapid progress of AI in mathematics is disruptive - Dan Roberts, OpenAI

A spokesperson for the Mathathon organizers said that despite losing one of its main sponsors, the event will go ahead largely unchanged, since talks are already underway with other companies willing to put up a comparable amount per team. The organizers stressed they do not expect OpenAI's withdrawal to significantly affect the competition.

Background: Navier-Stokes and Buckmaster's allegations

Tensions grew after OpenAI announced it had solved the Navier-Stokes problem, one of the seven Millennium Prize Problems. Tristan Buckmaster, a professor at New York University, had earlier accused OpenAI of pressuring him not to give due credit to a collaborator affiliated with Anthropic, and suggested the company may have used their joint work, conducted with Codex, to produce its own proof during a weekend-long computing marathon. OpenAI's proof itself remains unverified by independent mathematicians.

The letter from the 25 Fields Medal laureates also references the earlier Leiden Declaration, published in June 2026 at leidendeclaration.ai, in which the mathematics community warned of a threat to values central to scientific work: verifiable proof, proper attribution of authorship, and independence from commercial interests.

What it means for the industry

The episode shows that the conflict over AI labs' working methods is shifting from individual disputes over authorship to an institutional level, where companies are losing partnerships and sponsorships under pressure from the scientific community. For companies like OpenAI and Anthropic, which like to present solving difficult math problems as proof of their models' progress, public opposition from Fields Medal laureates carries a real reputational cost, regardless of whether individual allegations of appropriating work are ever confirmed.

For Polish researchers and universities following these developments, the episode signals that authorship attribution for work produced with the help of AI models will need clear rules before similar disputes arise locally. The growing number of signatories to documents such as the Leiden Declaration and the mathandai.org letter suggests the mathematics community wants to establish shared standards before the commercial race among AI labs fully shapes publication practices in the field.