OpenAI’s feud with mathematicians is only escalating
Twenty-five Fields Medalists have endorsed an open letter warning that competitive AI labs are endangering mathematicians' intellectual labor by racing to publish solutions to major problems. The letter follows specific disputes — including an NYU professor's claim that OpenAI pressured him over crediting an Anthropic-associated collaborator and the company's withdrawal of sponsorship from a Caltech event amid criticism.
Why It Matters
The dispute highlights how powerful AI models, and the incentives of well-funded labs, could reshape who produces and gets credit for mathematical advances — potentially undermining the collaborative, open verification processes that sustain the discipline.
Key Facts
- Signatories: 25 leading mathematicians
- Honors held by signatories: Each signatory has received a Fields Medal
- Accusation by Tristan Buckmaster: NYU professor said OpenAI pressured him not to credit an Anthropic-associated collaborator and questioned whether OpenAI used Codex-influenced work to produce its proof
- Event sponsorship withdrawn: OpenAI withdrew sponsorship of a Caltech math event after criticism from researchers
- Verification status: OpenAI's reported proof remains unverified
A group of 25 prominent mathematicians — all recipients of the Fields Medal — has issued an open letter raising alarms about how competition among AI developers is affecting mathematical research. The signatories argue that when labs rush to announce machine-generated solutions to major problems, there is little time for rigorous writeups, methodological clarification, or proper citation of prior work, leaving attribution and verification gaps.
The tensions have become personal and public. NYU professor Tristan Buckmaster accused OpenAI of pressuring him not to acknowledge a collaborator affiliated with Anthropic in connection with a significant proof, and said he suspected OpenAI might have leveraged Codex-influenced material to generate its own result during a prolonged inference run. Separately, OpenAI pulled its sponsorship of a math event at Caltech after researchers at the university criticized the company.
The letter stresses that mathematical advances are valuable not only as final proofs but as intellectual frameworks that train students, spawn new questions, and integrate findings into the broader literature. Without careful human-led development, the authors warn, AI-produced ideas may not be fully assimilated into the discipline or transmitted through the community's traditional channels.
Signers point to a larger structural risk: well-funded labs can deploy large language model resources at great expense and speed to pursue discoveries, a dynamic that could incentivize secrecy over open collaboration. The statement follows the Leiden Declaration released in June by a working group of mathematicians, which similarly considered how LLM-generated proofs will change research practices and recommended steps for scholars and institutions.
While the letter acknowledges the potential benefits of AI solving hard mathematical problems, it insists those benefits depend on transparent, verifiable, and properly credited work carried out in partnership with the mathematical community. The authors also note that the concerns now surfacing in mathematics mirror challenges other scientific and creative fields may soon face as AI alters how work is produced and disseminated.
Keep Reading
TikTok misled consumers about filtering content for minors, judge rules
Nscale adds former OpenAI exec Fidji Simo to its board ahead of potential IPO
Roblox is making it easier to build games with AI — and play them outside Roblox
