OpenAI Says People Linked to China's Moonshot Tried to Copy Its AI's Hidden Reasoning
OpenAI says it disrupted a coordinated effort to extract and reproduce hidden reasoning traces from its models, tracing a core cluster to people associated with Moonshot AI, the Chinese startup behind Kimi. The activity began July 1, spiked on July 24–25, and was fully blocked by July 28. OpenAI says the operators did not hack systems or steal chats; instead, they manipulated model interactions to expose protected internal reasoning in visible form. One tactic involved copying encrypted reasoning from one conversation and decoding it in another. OpenAI frames this as unauthorized “adversarial distillation,” where a weaker model is trained on a stronger model’s outputs or reasoning. The case fits a broader pattern of AI companies accusing rivals of model extraction and distillation, with similar disputes involving DeepSeek, Anthropic, xAI, and others.
