US 'Watermark' Found in Chinese AI Models? Treasury Secretary Bessent Vows Investigation and Possible Sanctions

US Treasury Secretary Scott Bessent stated on Tuesday (21st) that the US government will conduct strict reviews of open-source artificial intelligence (AI) models developed overseas, including products launched by Chinese companies. He emphasized that Washington has the capability to impose sanctions if evidence emerges that these models have stolen intellectual property from American enterprises.

Speaking in an interview with Fox Business, Bessent highlighted that the United States is a key hub for AI innovation and that the Trump administration supports open-source models, but will not tolerate intellectual property theft. He warned that if evidence shows overseas AI models have stolen technology or achievements from leading American companies, the US could take sanction measures in response to such infringement.

He further claimed that the US government has already detected traces of 'watermarks' from US large language models within many Chinese AI models—an unacceptable situation, in his view. Washington will conduct further examination of the evidence in the coming days and weeks, though he did not disclose which Chinese models or US companies are involved, nor did he specify the potential sanctions that might be imposed.

Bessent also indicated that the US government may evaluate whether companies using Chinese AI models should be required to disclose this information to their customers. He compared the issue to counterfeit goods, stressing that businesses cannot ethically use products built on stolen intellectual property.

This statement signals that the Trump administration’s scrutiny of China’s AI industry may expand beyond export controls on advanced chips to include model training data, technology origins, corporate disclosure obligations, and intellectual property disputes—adding new layers of regulatory and sanction risks to the US-China AI competition.