U.S. technology officials on July 22 said they have evidence that Chinese start-up Moonshot AI used Anthropic’s most advanced large language model, Fable, as the basis for its own K3 release and that the firm obtained high-end Nvidia hardware for model training. The remarks were posted publicly by a senior U.S. official and were followed by a separate Treasury comment indicating possible trade sanctions.
In a post on X, top Trump administration tech official Michael Kratsios wrote, "We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model." He added that "Large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable."
Kratsios further stated that Moonshot had acquired servers equipped with powerful Nvidia AI chips called GB300s and that those machines were used in Thailand, "likely to train its AI models." A senior administration official earlier this year also said that another Chinese start-up, DeepSeek, trained its latest model on Nvidia’s most advanced AI chip, a move that could represent a breach of U.S. export controls.
In a separate post on X, U.S. Treasury Secretary Scott Bessent said he was weighing the addition of Moonshot AI to a trade blacklist and imposing sanctions.
Responses and denials
Moonshot did not immediately respond to a request for comment on the U.S. allegations. A Chinese embassy spokesperson in Washington, Liu Chang, called Kratsios' remarks "entirely unfounded" and said China respects intellectual property protections. "Relevant individuals in the United States should respect the facts, discard prejudice and stop smearing and discrediting China’s achievements in the development of its artificial intelligence industry," Liu said.
Anthropic, meanwhile, voiced concern about Chinese firms' use of U.S. models. Anthropic public policy executive Sarah Heck wrote on X that Chinese theft of U.S. models "creates serious national security risks for the United States."
What distillation means and previous allegations
The administration's comments referenced "distillation," a process in which developers train smaller, lower-cost AI models using outputs from larger, more resource-intensive systems. The U.S. officials framed such large-scale, covert distillation efforts as a way for foreign actors to reproduce capabilities without conducting equivalent original research.
Anthropic has previously accused three Chinese AI companies - DeepSeek, Moonshot and MiniMax - of using its Claude model improperly. In February, Anthropic said those firms created more than 16 million interactions with Claude using roughly 24,000 fake accounts, in violation of Anthropic’s terms of service and regional access restrictions.
OpenAI has also warned U.S. lawmakers that DeepSeek targeted ChatGPT and leading U.S. AI companies in attempts to replicate their systems for use in its own training, according to prior reporting by government sources.
In April, the U.S. State Department initiated a global outreach effort to highlight what it described as widespread efforts by Chinese companies, including Moonshot, to obtain intellectual property from U.S. AI labs. A diplomatic cable noted that, "AI models developed from surreptitious, unauthorized distillation campaigns enable foreign actors to release products that appear to perform comparably on select benchmarks at a fraction of the cost but do not replicate the full performance of the original system."
Moonshot's K3 and the evolving AI landscape
Moonshot on Friday unveiled Kimi K3, a model the company said contains 2.8 trillion parameters and represents the world's largest open-weight AI system, claiming performance that approaches Anthropic's frontier Fable model. The launch follows the abrupt withdrawal by U.S. authorities of Anthropic’s Fable and Mythos models last month for security reasons.
The U.S. accusations and potential trade actions come as firms on both sides of the Pacific roll out increasingly capable AI models with broad commercial uses and potential military implications. U.S. officials have pointed to hardware transfers and unauthorized model replication as key concerns in safeguarding proprietary technology and research.
Implications
The sequence of allegations, denials and potential sanctions underscores rising friction around advanced AI capabilities and the movement of specialized AI hardware. U.S. officials are framing the issue as both an economic and a national security concern, while Chinese officials reject what they describe as politically motivated attacks on their technological progress.
Given the public postings by senior U.S. figures and the Treasury's consideration of trade restrictions, the dispute is likely to factor into broader policy decisions about export controls, cross-border hardware transfers and how governments and companies protect AI training data and models.