U.S. Rep. Bill Huizenga has introduced legislation in the House to keep state-backed Chinese companies from using American models to build their own artificial intelligence platforms.
Huizenga, R-Holland Twp., on April 20 introduced the Deterring American AI Model Theft Act. The congressman says companies such as OpenAI, Google, and Anthropic have said publicly that they have faced “distillation” attacks from state-backed Chinese AI companies. These attacks systematically extract and steal critical data, such as model weights, architectures, and other technical characteristics, that foreign adversaries use to build their own AI models for cheap while bypassing compute restrictions.
Distillation, when carried out with the approval of the AI model owner, is a legitimate and valuable tool for research and innovation. This bill protects such activity by disentangling it from illicit “model extraction attacks.”
“Model extraction attacks are the latest frontier of Chinese economic coercion and theft of US intellectual property,” Huizenga said. “American AI models are demonstrating transformative cyber capabilities, and it is critical we prevent China from stealing these technological advancements.”
The Deterring American AI Model Theft Act addresses this problem by authorizing discretionary sanctions, creating a name and shame list of AI thieves, and approving a pathway for the U.S. State Department to interphase with private industry to share best practices and study attacks to strengthen defense.
Huizenga, R-Holland Twp., on April 20 introduced the Deterring American AI Model Theft Act. The congressman says companies such as OpenAI, Google, and Anthropic have said publicly that they have faced “distillation” attacks from state-backed Chinese AI companies. These attacks systematically extract and steal critical data, such as model weights, architectures, and other technical characteristics, that foreign adversaries use to build their own AI models for cheap while bypassing compute restrictions.
Distillation, when carried out with the approval of the AI model owner, is a legitimate and valuable tool for research and innovation. This bill protects such activity by disentangling it from illicit “model extraction attacks.”
“Model extraction attacks are the latest frontier of Chinese economic coercion and theft of US intellectual property,” Huizenga said. “American AI models are demonstrating transformative cyber capabilities, and it is critical we prevent China from stealing these technological advancements.”
The Deterring American AI Model Theft Act addresses this problem by authorizing discretionary sanctions, creating a name and shame list of AI thieves, and approving a pathway for the U.S. State Department to interphase with private industry to share best practices and study attacks to strengthen defense.
