US Weighs Sanctions on Chinese AI Models Over Alleged IP Theft
Treasury Secretary Scott Bessent said Washington could punish Chinese open-weight developers if probes confirm they distilled U.S. models, days after Moonshot's Kimi K3 rattled American labs.
CHINA — The United States could sanction Chinese artificial intelligence companies if it confirms their open models rely on stolen U.S. technology, Treasury Secretary Scott Bessent said.
Bessent said the administration would examine high-profile Chinese open-weight models for signs of intellectual property theft and impose penalties if it finds them, according to techcrunch. The threat marks the first time Washington has signaled it may target Chinese model developers directly. For years it has instead restricted China’s access to advanced chips and tightened export controls.
“This administration supports open source models, but what we do not support is IP theft,” Bessent said on Fox Business on July 21, 2026. “If we see, especially, that overseas models are stealing from our great companies, we have the ability to sanction them because of this theft.”
The warning lands as Chinese systems close the performance gap with American frontier labs. Moonshot AI, one of China’s leading developers, unveiled its Kimi K3 model days earlier, and independent testing from Artificial Analysis and Arena.ai found it competitive with Anthropic’s Fable 5 and OpenAI’s GPT-5.6, according to gizmodo.com. Moonshot priced Kimi K3 at $15 per million output tokens, against roughly $30 for GPT-5.6 Sol and $50 for Fable 5, according to theverge.
The fight centers on model distillation, a technique that trains a smaller, cheaper system on another model’s outputs. Anthropic accused operators tied to Alibaba of running roughly 25,000 fraudulent accounts to generate more than 28.8 million interactions with its Claude models between April 22 and June 5, according to pymnts.com. The company made similar claims in February against DeepSeek, Moonshot AI and MiniMax.
Not everyone accepts that distillation amounts to theft. Microsoft Chief Executive Satya Nadella called it “ironic” for AI companies that rely on fair use to train on public data to then restrict distillation. Hugging Face Chief Executive Clem Delangue went further, telling TechCrunch’s Equity podcast that distillation is “a very small factor” and a practice “everyone is doing, including companies in the U.S.,” crediting China’s progress to strong research teams and a more open approach.
The stakes reach beyond the labs. Cheaper, capable Chinese models threaten the business plans of OpenAI and Anthropic and their ability to raise capital for costly frontier development, according to siliconangle.com. American businesses increasingly turn to Chinese tools as domestic pricing climbs, according to theverge.
This is not the first American reaction to a Chinese AI leap. DeepSeek’s R1, released in January 2025, matched OpenAI’s best model on some tasks without the most expensive Nvidia chips. That launch triggered a trillion-dollar market selloff and pushed the administration toward export bans, according to gizmodo.com. In April, the White House vowed to help U.S. labs identify and thwart adversarial distillation.
The dispute has split Trump’s own advisers. David Sacks, the president’s former AI and crypto czar, criticized top labs that “want the government to eliminate their open source competition,” while Pentagon official Emil Michael and former adviser Dean Ball traded insults over whether Washington should quietly pressure companies away from models like Kimi, according to technologyreview.
A decision date for any sanctions has not been announced. Bessent said the administration would look into the matter in “the coming days or weeks,” and Reuters reported the U.S. and China plan their first bilateral AI talks in September 2026, led by Bessent, according to gizmodo.com.
Washington has reached for the language of a technological arms race before. When the Soviet Union launched Sputnik in October 1957, the shock spurred the U.S. to found NASA in 1958 and pour money into science and space programs, according to gizmodo.com. Commentators have repeatedly invoked that moment for Chinese AI, first with DeepSeek and now with Kimi, though what followed Sputnik was sustained state investment rather than a single punitive strike.