China Says US AI Firms Distilled Chinese Models, Cites No Evidence

China’s Ministry of Commerce on Monday accused “many American AI enterprises” of distilling Chinese models and promised “all necessary measures” if the Trump administration sanctions Chinese firms over alleged intellectual-property theft. The statement rejected accusations that Chinese companies stole US intellectual property through model distillation and asked Washington to end what it called a “smear campaign.” It followed Treasury Secretary Scott Bessent’s July 21 warning that the administration could impose sanctions.

Bessent’s July 21 sanctions threat

Bessent told Fox Business on July 21 that foreign models found to have stolen from American firms could face sanctions. He added that the administration had found watermarks from US large language models in many Chinese systems.

The ministry supplied no company names or evidence for its counter-accusation against American developers. “These actions lack factual basis and legal support, and constitute double standards in practice, representing typical acts of AI hegemony,” the ministry wrote in its statement.

The White House Office of Science and Technology Policy called Chinese distillation operations “deliberate, industrial-scale campaigns” in an April 23 memo.

Anthropic’s February report identified DeepSeek, MiniMax and Kimi K3 developer Moonshot AI. The company alleged that the three labs generated more than 16 million exchanges with Claude through roughly 24,000 fraudulent accounts. A Just Security analysis also described commercial proxy services managing 20,000 simultaneous fraudulent accounts and mixing extraction traffic with unrelated requests to evade detection.

Nathan Lambert of the Allen Institute for AI assessed in July that adversarial distillation contributed to Kimi K3 “at most to a relatively small degree.”

The July 24 open-weight letter

Beijing cited two American letters. Nearly 200 startups organized by the Little Tech Association, including Y Combinator and Proton, had asked the White House to preserve access to Chinese open-weight models. Their letter argued that a cutoff would hurt small companies while doing little to keep the technology out of users’ hands.

A separate July 24 letter, “Open Weights and American AI Leadership,” carried signatures from Nvidia, Microsoft, Meta, Palantir and 21 other companies. “Policymakers should be careful not to conflate legitimate model-development techniques with misappropriation,” the signers wrote. The document described distillation as widely used for model improvement, evaluation and validation. It directed unlawful-extraction concerns toward “targeted legal and commercial frameworks rather than sweeping restrictions.”

Kimi K3 at 1,679 points

The ministry also stated that Chinese models had reached leading capability in front-end coding. Arena’s Frontend Code leaderboard placed Kimi K3 first at 1,679 points, followed by Anthropic’s Claude Fable 5 at 1,631 and OpenAI’s GPT-5.6 Sol at 1,618. K3 led six of the board’s seven front-end categories, while Fable 5 ranked first in Gaming.

Tom’s Hardware cautioned that the benchmarked results came through API access and could not be verified independently at the time. Moonshot had previewed K3 on July 16 as a 2.8-trillion-parameter open-weight mixture-of-experts model and announced plans to publish the full weights on July 27.

China’s next export catalogue

The Financial Times reported on July 21 that the Ministry of Commerce had begun consulting domestic industry about possible export controls for model weights, key training data and Chinese chip designs. The proposals could enter the next revision of China’s catalogue of technologies prohibited or restricted from export. Reuters said it could not immediately verify the report, and the ministry has not published a revised catalogue containing those items.

Under rules issued with the 2025 catalogue, China bars exports of prohibited technologies and requires licenses for restricted ones. Technologies outside the catalogue require contract registration.

US and Chinese officials plan to hold AI talks in September, Reuters reported, probably before President Xi Jinping’s planned Sept. 24 visit to the United States. Bessent is expected to lead the American side, and dates for the talks have not been finalized.

Jensen Huang Defends Chinese AI Models Hours After Bessent Sanctions Threat

Nvidia CEO Jensen Huang told Axios on Tuesday that American companies should "absolutely" be allowed to use Chinese AI models. The remarks came hours after Treasury Secretary Scott Bessent threatened

The Implicator

OpenAI and Anthropic Lobby Washington to Restrict Chinese Open-Weight AI

OpenAI and Anthropic have lobbied Washington regulators to restrict Chinese open-weight AI models, five people close to the discussions told The New York Times on Saturday. According to the same repor

The Implicator

White House Accuses Moonshot of Distilling Fable and Using Banned Nvidia Chips

Michael Kratsios, director of the White House Office of Science and Technology Policy, accused Moonshot AI on Wednesday of distilling Anthropic’s Fable to develop Kimi K3, the 2.8-trillion-parameter m

The Implicator

Marcus Schuler

San Francisco

Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai