OpenAI Links Large-Scale Reasoning Extraction Campaign to Moonshot AI

OpenAI said it disrupted a coordinated reasoning-extraction campaign that began July 1 and surged July 24–25 with 16,000 extraction requests from more than 4,000 users, attributing a core group of the activity to individuals associated with Chinese developer Moonshot AI. OpenAI identified related activity across more than 15,000 users and disrupted it by July 28, restricting accounts and tightening technical safeguards; the company said the campaign neither broke its encryption nor compromised databases. The allegations add to scrutiny of model distillation, which Anthropic has also reported against Claude, with activity it attributed to Moonshot, Alibaba and DeepSeek.

OpenAI said Wednesday it disrupted a coordinated effort to extract protected reasoning from its models, attributing a core group of the activity to individuals associated with Moonshot AI, the Chinese developer of the Kimi chatbot. The company said activity began July 1 and surged July 24–25, with 16,000 extraction requests from more than 4,000 users. Investigators subsequently identified related activity across more than 15,000 users, which OpenAI said it disrupted by July 28. Those figures represent attempted extractions, not necessarily successful ones. Operators allegedly copied encrypted reasoning between conversations and prompted models to reveal its contents. OpenAI said the campaign neither broke its encryption nor compromised databases. It remains uncertain whether every operator belonged to the same organization. The company responded by restricting accounts, tightening technical safeguards and coordinating with industry partners. It warned that extracting reasoning could help competitors reproduce capabilities without retaining the original models’ safety protections. The allegations add to mounting scrutiny of model distillation, in which one AI system’s responses are used to improve another. Anthropic has separately reported campaigns targeting Claude, including activity it attributed to Moonshot, Alibaba and DeepSeek. For AI developers and investors, the dispute raises a central competitive question: how effectively can companies protect capabilities developed through costly research and computing investments when rivals can repeatedly query their products?