Treasury Secretary says it's 'unacceptable' that Chinese open source AI models may have been built on 'IP theft'
Treasury Secretary Scott Bessent recently signaled that the United States government is prepared to take action against Chinese artificial intelligence firms suspected of intellectual property theft. The administration is currently investigating whether prominent Chinese open-source models rely on stolen data from domestic competitors. While officials maintain that they support the open-source movement, they draw a hard line when that output is built upon proprietary research obtained without authorization.
During a recent interview, Bessent described the technical practice of distillation as theft when it involves unauthorized access to American technology. The administration claims to have identified specific watermarks from United States large language models within several Chinese systems. Because of these findings, the Treasury Department is weighing the use of sanctions to protect domestic innovation and maintain a level playing field for companies like Anthropic, OpenAI, and other industry leaders.
This development comes as models like Moonshot AI's Kimi K3 gain significant traction, performing at or above the capability levels of top Western proprietary software. The rise of these tools has ignited a debate within the tech industry. Some investors argue that open-source alternatives are necessary to prevent a market duopoly, while others worry about the security implications and the legitimacy of the training data used to create these powerful systems.
Bessent emphasized that while US entities like Meta and Nvidia continue to push their own open-source projects, those efforts must adhere to strict standards. The government intends to ensure that no foreign entity gains a market advantage by bypassing intellectual property protections. As the deadline for the public release of Kimi 3 approaches, the tension between international innovation and national security remains a primary focus for Washington.

