Treasury Secretary Threatens Sanctions Over Chinese AI 'Distillation'

U.S. Weighs Sanctions Against Chinese AI Companies

U.S. Treasury Secretary Scott Bessent announced Tuesday that the Trump administration will investigate whether Chinese artificial intelligence companies have illegally copied American AI models through a process called distillation, warning that sanctions could follow if theft is confirmed.

"If we see, especially that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft," Bessent said in an interview with Fox Business. He added that investigators have found "watermarks of our U.S. large language models on many of the Chinese models, and that's unacceptable."

The comments come as Chinese open-weight AI models gain competitive ground against leading American offerings from OpenAI and Anthropic. Earlier this month, Chinese startup Moonshot AI released Kimi K3, a model that outperforms some U.S. competitors on industry benchmarks.

The Distillation Controversy

Distillation is an AI training technique where developers build a smaller, less capable model using outputs from an existing, more powerful system. While the method has legitimate applications, it can also enable unauthorized replication of proprietary technology.

Last month, Anthropic sent a letter to the U.S. Senate Committee on Banking, Housing, and Urban Affairs alleging that Chinese tech giant Alibaba conducted "the largest known distillation attack" against its models to date.

Bessent indicated the administration would examine the evidence "in the coming days or weeks" and determine appropriate responses.

Bilateral Talks Scheduled

The investigation unfolds as the U.S. and China prepare to hold AI-focused discussions in September, with Bessent representing the American side, according to a Reuters report published Tuesday.

The timing suggests the distillation allegations could become a central point of contention in the talks, as both nations seek to establish norms around AI development and intellectual property protection.

Why It Matters

The threat of sanctions represents a significant escalation in U.S. efforts to protect its AI advantage, moving beyond export controls on chips to directly targeting alleged intellectual property violations. However, the move could invite scrutiny of American companies' own training practices—both Anthropic and OpenAI have faced legal challenges over their use of copyrighted material. Anthropic settled a $1.5 billion class action lawsuit with authors in September over allegedly using pirated books, while The New York Times sued OpenAI and Microsoft in 2023 for copyright infringement. The distillation dispute highlights the absence of clear international standards for AI development as the technology reshapes global competition.

These details were first reported by CNBC.