The US is gearing up to tackle what it views as industrial-scale theft of intellectual property from American AI labs by Chinese entities.

Since the emergence of DeepSeek—a Chinese model that prompted OpenAI to claim its own models were used in training—it has been alleged that other firms have employed a technique called distillation to steal proprietary information. In January, Google warned of 'commercially motivated' actors attempting to clone its Gemini AI chatbot. Anthropic followed suit, accusing Chinese firms of using the same method to generate exchanges with Claude.

Michael Kratsios, director of the White House Office of Science and Technology Policy, has indicated that foreign entities in China are involved in deliberate campaigns to distill US frontier AI systems. The memo suggests these attacks use tens of thousands of proxy accounts to evade detection and jailbreaking techniques to expose proprietary information.

US firms have so far alleged such attacks violate their terms of service, but Congress may soon update laws to better equip companies defending against the alleged fraud. Kratsios has confirmed that measures are being explored to hold foreign actors accountable for these campaigns.