The US government's intensified focus on countering foreign artificial intelligence threats, particularly from China, marks a significant shift in its strategic approach. A recent memo from the White House outlines concerns regarding foreign entities' efforts to replicate American AI systems through “industrial-scale campaigns.” This term refers to the process of distillation, where a model is trained using outputs from a more advanced system, mimicking its capabilities.
Accusations from AI firm Anthropic have highlighted the activities of companies like DeepSeek, Moonshot, and MiniMax, which are alleged to conduct “industrial-scale distillation attacks.” Discussions between Anthropic and the White House suggest the company influenced the government's response to these threats. The White House plans to share relevant information with American AI firms and enhance coordination to safeguard technological advancements.
Additionally, Michael Kratsios, a leading science and technology advisor, indicated that Chinese entities are employing tactics such as using “tens of thousands of proxy accounts” and jailbreaking techniques to access proprietary information. This multi-layered strategy has prompted the US to consider holding foreign actors accountable. Meanwhile, the Chinese embassy in Washington has rejected these allegations, labeling them as “pure slander” and emphasizing China's commitment to protecting intellectual property rights.