NYU professor Tristan Buckmaster, a Fields Medal recipient, has accused OpenAI of pressuring him not to credit a collaborator who works for Anthropic for solving an important math problem, and questioned whether OpenAI used their Codex work to produce its own proof over a marathon inference weekend. OpenAI subsequently withdrew its sponsorship of a math event at CalTech after facing criticism from university researchers.
A group of mathematicians argue that while AI's ability to solve outstanding mathematical challenges could benefit humanity, those solutions must be understood and communicated by the math community. They wrote that solutions are often announced hastily, leaving no time for proper documentation or citing previous work, and that OpenAI's proof remains unverified. The mathematicians raise attribution and plagiarism concerns, warning that without willing mathematicians to integrate AI-conceived ideas into the mathematical canon, the crucial human transmission chain would be lost.
Other mathematicians are growing concerned that their use of Codex may have been fed into OpenAI's new models, threatening the culture of open research. Industry observers note that frontier labs can now spend tens of millions of dollars using LLMs to beat original researchers to a proof—a dynamic that will incentivize secrecy. The letter follows the Leiden Declaration released by mathematicians in June, which offered recommendations on how LLM proofs will change mathematical work. The mathematicians warn that other scientific and creative professions face similar issues and "your field of interest is next."