A cluster of users associated with Moonshot AI, the developer of the Kimi chatbot, targeted protected reasoning within OpenAI models starting July 1. The company identified over 15,000 users involved in the coordinated effort to distill its models, with attempts peaking at 16,000 across more than 4,000 accounts on July 24 and 25. OpenAI banned or restricted the accounts and closed the extraction pathway to stop the activity.

No internal databases or encryption were compromised, and no stored user conversations were accessed during the campaign. One identified method involved extracting encrypted reasoning from one conversation and using a separate model to decrypt and reproduce the content. OpenAI noted that the reported figures represent attempted rather than successful extractions and has shared its findings with government and industry partners.