Mathematician suspects OpenAI used his unpublished work for a proof

Australian mathematician Tristan Buckmaster of NYU uploaded drafts of his Navier–Stokes research into OpenAI's Codex tool. Four days later an unreleased OpenAI model produced a possible proof of a Millennium Prize problem, and he suspected the AI drew on his unpublished work. OpenAI denies the influence but admits user data may be used to train its models.

• Buckmaster studies the Navier–Stokes equations, one of seven Millennium Prize Problems
• An OpenAI model produced a possible proof four days after a September 1 request
• OpenAI denies using the work but allows user data for model training
• Indian institutes IIIT Hyderabad and MAHE are drafting AI research policies