The US State Department has launched a global diplomatic effort to address concerns that Chinese AI firms, including Moonshot AI, DeepSeek, and MiniMax, are illicitly distilling US AI models to train their own systems.
Goldman Sachs has barred its bankers in Hong Kong from using Anthropic's Claude AI models, following consultations with Anthropic and adopting a strict interpretation of its contract. The restriction does not extend to other AI vendors like OpenAI. Western AI models are banned in mainland China under the Great Firewall, while Hong Kong has generally operated outside those restrictions. US AI firms are concerned about 'distillation' - the risk that intensive use in China could help local actors train competing systems. The move could pose challenges for Hong Kong's role as a financial hub as bankers may fall behind peers elsewhere.
Goldman Sachs has prohibited its Hong Kong bankers from using Anthropic's Claude AI models, citing contract interpretations amid US-China tech tensions. This move, impacting coding and financial modeling, raises concerns for Hong Kong's financial hub status and other institutions using the AI. The restriction highlights broader worries about AI intellectual property theft and cybersecurity risks.
The Trump administration announced plans to combat foreign tech companies' exploitation of U.S. AI models, specifically targeting China. A Thursday memo from Michael Kratsios, the president's chief science and technology adviser, accused Chinese entities of industrial-scale campaigns to 'distill' capabilities from leading U.S. AI systems. The administration will work with American AI companies to identify such activities, build defenses, and punish offenders. The memo comes as China's AI performance gap with the U.S. has 'effectively closed' according to Stanford researchers, and as the House Foreign Affairs Committee advances bipartisan legislation to sanction foreign actors who extract key technical features from U.S. AI models.
The US is preparing to crack down on China's alleged industrial-scale theft of American AI intellectual property through distillation attacks. White House OSTP Director Michael Kratsios warned in a memo that foreign entities principally based in China are using tens of thousands of proxy accounts to distill US frontier AI systems. Google, Anthropic, and OpenAI have all accused Chinese firms of using distillation to clone their models. China has dismissed the accusations as pure slander, while Congress considers treating model extraction as industrial espionage.
