Abstract:
According to Reuters, the U.S. government on Tuesday accused Chinese artificial intelligence companies of maliciously copying the technology of American AI companies. U.S. law enforcement and intelligence officials said the Chinese companies copied intellectual property from U.S. AI labs through a technique called "distillation." The so-called "distillation" refers to using larger and more expensive model output to train smaller AI models to reduce the training cost of new tools.
U.S. officials said that a total of six Chinese companies were named, including DeepSeek, Moonshot AI, and Alibaba. They used different variants of the U.S. AI model to speed up the training of their own products. Officials also said these practices were "likely with the knowledge of the Chinese government."
The United States said that the American AI models targeted by these companies come from Anthropic, OpenAI, Google and SpaceX. Officials also said that China's homegrown AI companies are engaged in "aggressive, malicious and targeted distillation activities" and have reached an "industrial scale".
The accusation comes as the Trump administration prepares for Chinese President Xi Jinping’s planned visit to the United States in late September. The United States and China are also preparing for a dialogue in mid-September to discuss AI security risks.
The Chinese Embassy in Washington did not immediately respond to a Reuters request for comment. Reuters also pointed out that the United States made similar accusations before Trump's visit to Beijing in April this year.
The U.S. statement on Tuesday also said that this "distillation" will not only reduce the research and development costs of Chinese AI companies, but also enhance China's "military and cyber attack capabilities," which may be used against the United States and its allies. Reuters also previously reported that Chinese military researchers have used the output of leading American AI models to train domestic AI systems and improve defense capabilities.
Comments