Scott Bessent Threatens Sanctions on Chinese Open-Source AI Models Over IP Theft

U.S. Treasury Secretary Scott Bessent said on Tuesday that the Trump administration will look into whether Chinese open-source AI models are stealing intellectual property from American companies and that it has “the ability to sanction.”

In remarks tied to Fox Business, Bessent said, “This administration supports open source models, but what we do not support is IP theft,” and he warned that overseas models “are stealing from our great companies.”

Ars Technica

Bessent also said the administration is finding “watermarks of our U.S. large language models on many of the Chinese models,” and that it will be “looking into the matter in ‘the coming days or weeks.’”

The dispute is framed around whether Chinese developers used “distillation,” described as a training method where one model’s outputs are used to train another, to replicate capabilities from U.S. frontier systems.

Bessent’s threat centers on detection of U.S. model “watermarks” on Chinese systems, with MeriTalk quoting him saying, “So, we’re going to be looking at that in the coming days or weeks.”

The same reporting ties the accusation to “distillation,” with Bessent saying, “They have hundreds of millions of hits against the large language models here in the U.S. and they try to reconstruct their code and the thought process of the models, and that’s unacceptable.”

Bolsamania

Other coverage highlights the controversy over whether distillation is theft, with Geeknetic quoting Hugging Face’s Clem Delangue saying, “es una práctica extendida también entre compañías estadounidenses y representa solo una parte del proceso.”

In that framing, the debate is not only about sanctions but also about what counts as unauthorized copying, as the sources describe distillation as a way to transfer capabilities from a larger model to a smaller one.

The sanctions warning arrives as the United States and China plan their first bilateral AI dialogue under President Donald Trump, with Reuters reporting it will take place in September and be led by Bessent on the U.S. side.

“Afirma que hay marcas de agua de modelos estadounidenses en IA chinas Bolsamania Tiempo de lectura:3min El secretario del Tesoro de Estados Unidos, Scott Bessent, ha afirmado este martes que el Gobierno está investigando simodelos chinos de inteligencia artificialhan sido entrenados mediante prácticas que suponen unrobo de propiedad intelectuala modelos estadounidense, lo que podría desembocar en sanciones”

Bolsamania

Gizmodo adds that Bessent is expected to raise AI distillation concerns in those talks, while Chinese authorities are expected to raise concerns about Anthropic’s “Mythos,” which the source says can allegedly hack into even the most secure systems.

The sources also connect the pressure campaign to market and corporate stakes, with Gizmodo describing how Chinese open-weight models like Moonshot AI’s Kimi K3 and Alibaba-backed systems are challenging U.S. leadership and prompting export bans and stricter trade restrictions.

In parallel, the EU threat is framed as an extension of the same IP theft dispute, with noticiasneo stating that “La administración examina prácticas de código abierto chino” and that sanctions could follow if IP theft is confirmed.