OpenAI accused of scooping research via user chats, sparking priority debate

generatorman_ai · x · 2026-09-08

A thread debates whether OpenAI leveraged independent researcher Tristan's shared work to prompt its model in the same direction — a very different situation from parallel research, one side argues. A counterpoint: there's no evidence Tristan shared details with OpenAI, and his actual conjecture is that OpenAI trained on his chats, though training on days-old chats seems unlikely. The thread also disputes the notion that "getting scooped is normal" and researchers may simply ignore prior work and publish anyway.

Related event: OpenAI-Mathematicians Navier-Stokes Priority Dispute Escalates as Bubeck Denies Accusations(18 posts)→

Original post →

More from Fun

Fun channel →