The White House is moving to criminalize AI model distillation attacks targeting US labs. Michael Kratsios, director of the Office of Science and Technology Policy, wrote in a memo that foreign entities, principally China-based, are running deliberate, industrial-scale campaigns to distill US frontier AI systems.
The evidence is specific. Google logged over 100,000 prompts from commercially motivated actors attempting to clone Gemini. Anthropic traced 16 million Claude exchanges across roughly 24,000 fraudulent accounts back to DeepSeek, Moonshot, and MiniMax. OpenAI confirmed the majority of distillation attacks it observed originated from China, following its earlier accusation that DeepSeek trained on its model outputs.
Read the full piece for China's formal response and the policy mechanisms the US is considering. The legal framing of distillation as theft, not just terms-of-service violation, is the detail that matters most going forward.
[READ ORIGINAL →]