A post by mr. TIM, dated September 9, 2026, relays a researcher's accusation that OpenAI engaged in "training on conversations" and then presented the result as a breakthrough. Published via timkellogg.me and timestamped 22:51 UTC, the brief post identifies no researcher, cites no evidence, and includes no response from OpenAI. The accusation, categorized as a rumor, remains unverified.
No score is assigned. Sources and their independence are shown in the citation chain below.