Twenty-five distinguished mathematicians, all recipients of the Fields Medal, have publicly expressed concerns about the impact of AI labs on their field. They argue that the rush to use AI to solve famous mathematical problems risks undermining the collaborative and transparent nature of mathematical research. Recently, NYU professor Tristan Buckmaster accused OpenAI of pressuring him to omit credit for a collaborator affiliated with Anthropic in relation to a significant mathematical breakthrough. Buckmaster also questioned whether OpenAI leveraged its Codex model to independently generate a major proof during an intensive inference session.
In response to criticism from CalTech researchers, OpenAI withdrew its sponsorship of a mathematics event at the institution. The mathematicians emphasize that while AI’s ability to tackle complex problems could benefit humanity, such solutions must be properly documented, understood, and shared within the math community. They note that hastily announced AI-generated proofs often lack thorough write-ups, proper isolation of new methods, and appropriate citation of prior work. This raises concerns about attribution and potential plagiarism.
The letter highlights that without active involvement from mathematicians to develop and integrate AI-generated ideas, these contributions risk losing their significance and the essential human element of knowledge transmission. There is also growing unease among mathematicians about whether their interactions with AI tools like Codex are being used to train newer models, fueling fears that the open research culture may give way to secrecy as labs invest heavily to outpace traditional researchers.
This development follows the Leiden Declaration issued earlier this year, which addressed similar challenges posed by AI in mathematics and proposed recommendations for the community and policymakers. The mathematicians stress that the true value of their work lies not only in proofs but in the intellectual framework that fosters education, innovation, and integration into broader knowledge.
They caution that these concerns extend beyond mathematics, signaling challenges that other scientific and creative fields will face as AI reshapes workflows. The core issue is ensuring that as AI changes how work is done, the original purpose and meaning of that work remain clear and preserved.