Bessent says U.S. could sanction China over AI model 'theft' (www.cnbc.com)

U.S. Treasury Secretary Scott Bessent announced that the Trump administration is investigating claims that Chinese AI models have been derived from American models through a process known as distillation, where a less capable model learns from a more powerful one. This scrutiny comes as Chinese open-weight models, such as Moonshot AI's Kimi K3, gain traction against established American technologies from companies like OpenAI and Anthropic, raising alarms about the U.S.'s competitive edge in the AI landscape. Bessent emphasized that if evidence of intellectual property theft emerges, the U.S. has the means to impose sanctions on the responsible entities. The discussion highlights critical concerns within the AI/ML community regarding intellectual property rights and the methods companies use to train their models. Amid ongoing tensions, the U.S. plans to hold talks with China in September to address AI-related issues, with Bessent representing U.S. interests. He pointed out that the presence of U.S. model "watermarks" in Chinese products is particularly troubling, signaling potential infringement. This situation not only underscores the challenges of safeguarding intellectual property in AI but also reflects broader geopolitical dynamics that could shape the future of AI development and collaboration.