Math professor accuses OpenAI of racing to claim Navier-Stokes proof after learning of rival's progress
NYU professor Tristan Buckmaster announced preliminary proofs on the Navier-Stokes problem with Anthropic's Levent Alpöge, but alleges OpenAI obtained information about their work and used massive compute to attempt a full proof first. OpenAI's Sebastian Bubeck denies the claims, calling them false and inflammatory. The dispute involves a $1 million prize for solving the problem.
The disputed approach targets a specific path within the Navier-Stokes problem, known as the smooth force route, which the duo says almost no other researchers were pursuing. Alpöge, though employed by Anthropic, worked independently, and the pair relied heavily on OpenAI's Codex model for their computations.
Buckmaster alleges Bubeck proposed dropping Alpöge's credit as a compromise and made threatening remarks about career damage. He also worries that OpenAI's policy of training on Codex interactions may have allowed his private work to inform their rival attempt.
This dispute could reshape how mathematicians collaborate with AI labs, as fears of data leakage may discourage open sharing of preliminary results. If the allegations hold, it may undermine trust in OpenAI's commitment to academic norms, potentially slowing cross-institutional research. Conversely, it could spur clearer guidelines for AI training data usage and credit attribution, affecting both academic institutions and commercial AI developers.