United States intelligence and law enforcement agencies have formally identified six Chinese artificial intelligence companies accused of conducting large-scale distillation of American frontier AI models. In a joint statement released on Tuesday, the National Security Agency (NSA), the Cybersecurity and Infrastructure Security Agency (CISA), and the Federal Bureau of Investigation (FBI) named DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI as the entities involved.
According to the agencies, these firms have been targeting US-developed models—such as variants of Claude, GPT, Gemini, and Grok—since at least the latter part of 2024. The release stated that the companies likely operated with “Chinese government awareness” while extracting these capabilities.
The governments argued that this industrial-scale distillation offers significant strategic and economic advantages to the accused firms. By replicating the functionality of US models, Chinese companies can achieve substantially shorter development timelines and avoid the massive financial expenditures typically required to train frontier models from scratch.
Alibaba and DeepSeek? That is a surprising list. I wonder what specific distillation techniques they are actually using to pull this off.
This really underscores the strategic value of proprietary models. US tech giants need to stay ahead, or risk losing the edge completely.
Is the line between government support and company initiative really that distinct in this case? Hard to separate the two entirely.