The US government has accused six Chinese artificial intelligence companies of using a technique known as distillation to copy intellectual property from American AI labs. Officials claim this practice allows Chinese firms to train smaller models using output from larger, more expensive US-developed systems, thereby reducing their own research and development costs.
The companies identified by US authorities include DeepSeek, Moonshot AI, and Alibaba. According to the report, these entities have targeted AI models originally developed by OpenAI, Anthropic, Google, and SpaceX. US officials allege that these activities are conducted on an industrial scale and likely have the knowledge and support of the Chinese government.
US authorities warn that these distillation efforts could enhance China's military and cyber-attack capabilities, potentially posing a threat to the United States and its allies. The accusations come ahead of a scheduled dialogue on AI safety risks between the two nations, which is expected to take place in mid-September.
This is not the first time such allegations have been raised; the US government made similar claims in April. While the Chinese embassy in Washington has not yet provided a comment, the US continues to emphasize that Chinese military researchers have previously utilized output from advanced American AI models to improve domestic defense systems.
