The Trump administration signaled a tougher posture toward Chinese technology firms it says are siphoning capabilities from U.S. artificial intelligence models, according to a memo from Michael Kratsios, the president’s top science and technology adviser. The White House plans to coordinate with domestic AI companies to detect and deter “model extraction” and pursue penalties against offenders, as lawmakers advance a bipartisan House measure enabling sanctions on foreign actors who distill key features from closed-source U.S. systems. Beijing’s embassy rejected the allegations as unjustified suppression, while U.S. firms including OpenAI and Anthropic have accused Chinese labs of illicitly training rival models on outputs from their systems. The push comes as Stanford researchers say the performance gap between the countries’ leading AI models has largely closed, raising enforcement and attribution challenges that could complicate U.S.-China diplomacy ahead of an expected Trump-Xi meeting in mid-May.
Related articles:
– AI Safety Summit 2023: The Bletchley Declaration
– NIST AI Risk Management Framework 1.0
