On September 9, TechCrunch reported that Tristan Buckmaster, a mathematics professor at New York University, unveiled three proofs on Tuesday (local time). One of these preliminary results hints at a significant breakthrough regarding a long-standing unsolved problem in theoretical mathematics. He accused OpenAI of gaining insight into his research progress—while it was still unpublished—and utilizing its extensive computational capabilities to release a complete proof ahead of time. The controversy centers on the Navier-Stokes existence and smoothness problem, one of the seven 'Millennium Prize Problems' set forth by the Clay Mathematics Institute, which carries a $1 million reward.

Buckmaster alleged that as he and his collaborator, Levent Alpöge, were finalizing their research, they discovered that information about their progress had been shared with OpenAI. Their research predominantly relied on OpenAI's Codex model. However, Alpöge's association with Anthropic led to dissatisfaction from OpenAI. Buckmaster accused OpenAI of insisting on the removal of Alpöge's name from the research and raised concerns about whether OpenAI had used data from its interactions with Codex to train its models. OpenAI denied having access to unpublished research but acknowledged that it could not entirely dismiss the possibility that de-identified data had contributed to the enhancement of its models. The final proofs presented by both parties exhibited notable differences.