As China’s open-source large language models continue to shake up the US artificial intelligence industry, the US government is once again trying to crack down.
According to a report by Bloomberg, on July 21, US Treasury Secretary Bessent declared that the US will carefully review Chinese open-source AI models for any alleged “theft of US intellectual property” and will impose sanctions if such cases are found.
Speaking on Fox Business that day, Bessent threatened, “We’ve been keeping an eye on Chinese open-source models, and there’s a lot of talk about China’s AI rise threatening US large language models. But we do not support intellectual property theft. If we find that, especially if overseas models have stolen the work of great American companies, we have the ability to sanction those models.”
Bloomberg noted that last week, Chinese AI company Moonshot AI released its open-source model Kimi K3, which sparked widespread attention in the US tech community. Markets are worried that Chinese firms, with their lower costs and steadily improving models, are rapidly narrowing the gap with leading US companies—and could even undermine the competitive edge that Silicon Valley’s massive investments in AI infrastructure bring. These concerns sent US tech stocks tumbling last week.
The report added that besides the US Treasury Secretary, US AI companies like OpenAI and Anthropic have recently made repeated, baseless accusations that Chinese competitors are “using the outputs of leading US large language models in unauthorized ways to develop chatbots at a fraction of the cost.”
Bessent also jumped in to smear, claiming the US government has found “traces” of American large models in “many Chinese AI models,” calling this “unacceptable.”
“In the coming days or weeks, we’ll investigate this further,” he said.
As a key cabinet member in charge of fiscal matters, Bessent is deeply involved in shaping US AI and trade policy toward China. Reports say he has discussed AI security risks with banks and investment firms, and is pushing for an independent regulatory body to review the safety of AI models.
Beyond targeting Chinese AI models themselves, Bessent also suggested that US companies might be required to disclose whether they use Chinese AI models to their customers.
“Another issue could be whether companies using Chinese models should disclose that fact to their customers,” he said, labeling Chinese AI models as “counterfeit goods” and claiming, “You can’t use counterfeit goods.”
In another interview on July 21, US Trade Representative Jamieson Greer told CNBC that the Trump administration is assessing whether China is expanding its advantage in global AI competition through “unfair practices.” “We’re watching very closely how China is advancing its AI development, to ensure US companies can compete on a level playing field,” he said.
However, Gary Marcus, a prominent US AI scholar and professor emeritus at New York University, wrote on his personal blog that Chinese AI models have “almost caught up” with the US’s most advanced levels, and the US can’t “win” the AI race. He argued Washington should stop treating AI as a zero-sum game and instead explore international cooperation and public goods pathways.
He suggested, “If we could put AI back into the public domain, using international collaboration for medicine and science, that might be the most promising direction right now.”
Marcus also threw a challenge to Trump: “If Trump really wants a Nobel Peace Prize, this could be his chance—partnering with China, putting AI to use for the public good would be an incredible gift to humanity.”