US alleges Chinese AI firms misappropriated American models through distillation
Washington has accused six Chinese AI companies, including DeepSeek and Alibaba, of using distillation to copy intellectual property from US labs such as OpenAI and Anthropic. The allegation comes ahead of a planned meeting between President Xi and President Trump. US officials said the activity was likely carried out with Chinese government awareness.
The US accusation targets a training method where smaller systems learn from the outputs of larger, costlier models. Officials named DeepSeek, Moonshot AI, and Alibaba among six firms, asserting they improperly leveraged models from OpenAI, Anthropic, Google, and SpaceX.
The timing is notable, as these claims emerge just before a scheduled summit between Presidents Xi and Trump later this month. US law enforcement and intelligence sources indicated the copying was probably conducted with Beijing's awareness.
This accusation could intensify regulatory barriers for Chinese AI firms, potentially slowing their global expansion and raising costs for international businesses relying on their models. It may also heighten diplomatic friction, affecting broader tech trade and consumer access to advanced tools. For researchers and developers worldwide, the dispute could foster a more guarded environment, limiting open collaboration and innovation across borders.