U.S. intelligence and law enforcement officials have accused China of orchestrating the large-scale theft of intellectual property from American artificial intelligence laboratories through a technique called distillation. Distillation involves training smaller AI models using the outputs of larger, more expensive models, significantly reducing development costs. Officials allege six Chinese entities, DeepSeek, Moonshot AI, and Alibaba among them, are maliciously replicating U.S. AI technology from companies like OpenAI, Anthropic, Google, and SpaceX, “likely with Chinese government awareness.” The U.S. estimates American companies will spend roughly $700 billion this year on AI infrastructure, and contends this theft undermines that investment. The claims come as President Trump prepares to host Chinese President Xi Jinping and both countries schedule talks on AI safety risks. Chinese officials maintain their advancements are the result of “achieving high-level science and technological self-reliance” and urge the U.S. to avoid “false accusations.” The U.S. previously made similar allegations in April. Officials also state that this activity strengthens China’s military and cyberattack capabilities.
Distillation is not inherently malicious; it is a common research practice. Apple utilizes Google’s Gemini model through distillation to improve Siri, and university labs employ the technique for research purposes. However, U.S. officials assert the scale and intent of the Chinese activity are distinct, representing an industrial-scale effort to accelerate development and enhance strategic capabilities. Reports from July 31 indicate Chinese military researchers have already leveraged American AI outputs to improve domestic systems.
Read the original coverage
💬 Comments
📜 Comment Policy