OpenAI reported on September 30 that it disrupted a coordinated model-distillation campaign after activity that began July 1 surged on July 24 and 25. The company counted 16,000 requests using a relevant extraction pattern from more than 4,000 users and said related prompt patterns appeared across more than 15,000 users before disruption on July 28.
Attribution remains qualified
OpenAI attributes a core cluster to individuals associated with Moonshot AI, the developer of Kimi, but says it is unclear whether all observed operators came from one actor. The report concerns attempted extraction through model interactions; it does not describe a breach of OpenAI's encrypted systems, databases, or stored user conversations.
The attack class is independently documented
Independent researchers have shown ways to recover protected reasoning traces from proprietary model APIs, supporting the broader plausibility of reasoning-trace extraction. That paper does not validate OpenAI's campaign counts or attribution, which remain based on the company's telemetry and investigation.