The US Department of Commerce has issued new guidance aimed at further restricting China’s access to advanced artificial intelligence (AI) chips, closing what critics have described as regulatory loopholes that existed over the past year.
Under the new rules, Chinese companies’ subsidiaries located outside China, including those in countries such as Malaysia, will also be subject to US export licensing requirements for advanced semiconductors. These entities must obtain approval from the US government before purchasing high-end chips, in an effort to prevent Chinese firms from acquiring advanced semiconductor technology through third countries.
In recent years, the United States has increasingly relied on export controls to limit China’s access to critical semiconductors used in artificial intelligence, high-performance computing, and military technologies. However, the Department of Commerce’s decision last year not to enforce the so-called “AI Diffusion Rule” was widely seen as creating an unintended regulatory gap.
According to industry analysts, during the period when the policy was not strictly enforced, Chinese technology companies may have been able to purchase large quantities of advanced AI chips through overseas subsidiaries and use them for AI model training and related research and development activities. Industry sources estimate that hundreds of thousands of high-end chips may have been supplied to overseas entities linked to Chinese firms over the past year.
The Department of Commerce said the new guidance primarily applies to future exports and sales. It does not require existing data centres to stop using affected chips, nor does it prohibit companies from maintaining or servicing already deployed servers and high-performance computing equipment. As a result, advanced chips that have already reached overseas subsidiaries of Chinese companies can continue operating in the near term.