The Trump administration is reigniting its push to ban leading Chinese AI models in the United States, citing cybersecurity concerns following the recent release of Kimi K3 by Chinese startup Moonshot AI. According to an Axios report published July 20, downloadable open-weight models present an enforcement challenge that could make an outright U.S. ban nearly impossible.<\/p>\n
Chinese models such as DeepSeek and Kimi K3 are open-weight, meaning their trained parameters are published for public download. Enterprises can self-host these models on private infrastructure, keeping data in-house while reducing inference costs, factors driving adoption by U.S. companies.<\/p>\n
Citing sources close to the administration, the Axios report outlines several earlier actions the government weighed to curb Chinese AI expansion in the U.S. The Department of Commerce last year considered adding multiple Chinese AI labs, including DeepSeek, to its Entity List, a trade blacklist maintained by the Bureau of Industry and Security that restricts purchases of sensitive American hardware, software, or technology.<\/p>\n
Officials also drafted a joint advisory from the National Security Agency and the Office of the National Cyber Director to discourage use of Chinese AI models, and an executive order that would hold U.S. companies liable for security breaches involving hosted Chinese models. These efforts were initially paused amid internal pushback over market impacts but have been revived after the release of new Chinese open-weight models.<\/p>\n
Critics include former White House adviser Sriram Krishnan and David Sacks, an outside White House AI adviser. Sacks wrote on X on Sunday: “We are at a critical inflection point in AI policy. The leading closed labs, already a duopoly in terms of AI model revenue, want the government to eliminate their open-source competition.”<\/p>\n
The Axios report suggests that leading U.S. AI labs OpenAI and Anthropic may have a hand in the push for a ban, a framing that critics argue would entrench a duopoly and discourage competition.<\/p>\n
Chinese AI models are increasingly used by American companies due to their lower cost and capabilities perceived as comparable to domestic alternatives. Open-weight models like DeepSeek V4 and Kimi K3 allow companies to download and host them on private servers, providing data privacy and reducing API costs compared to closed Western alternatives.<\/p>\n
The pricing gap is significant. DeepSeek-V4-Pro charges $0.87 per million output tokens, compared with $50 for Anthropic’s frontier Claude Fable 5 model. Coinbase CEO Brian Armstrong noted that Coinbase runs models like GLM-5.2 and Kimi in production, cutting its overall AI spending nearly in half even as token consumption spiked.<\/p>\n
Self-hosting does shift costs, including GPUs, electricity, maintenance, networking, and model operations, onto the company, making the approach most economical for organizations with substantial and sustained AI usage.<\/p>\n
For individuals or small companies, using a model like DeepSeek via its website or app remains straightforward with a VPN, though app availability and payment restrictions still apply. Enterprise enforcement is far more complicated.<\/p>\n
Open-weight models exist as downloadable files mirrored across public repositories like Hugging Face and independent torrents, making them hard to recall once released. Once an American enterprise downloads the weights, it can run the model entirely offline inside a private, air-gapped data center, limiting regulators’ ability to monitor which model is running locally.<\/p>\n
Modifications add another layer of complexity. Companies routinely fine-tune, quantize, or distill these models, blending a Chinese base with domestic corporate data until provenance blurs and it becomes difficult to define where the foreign model ends and a new domestic one begins. Even under strict download bans, firms could host the models through subsidiaries, though that path runs into know-your-customer rules at cloud providers and the extraterritorial reach of U.S. export controls.<\/p>\n
According to Axios-cited government sources, the administration may not need an outright ban. The strategy appears to focus on getting U.S. firms to drop the models through procurement rules, Entity List threats, and public pressure campaigns aimed at companies using Chinese models. Sources also say the government will push to highlight potential backdoors and governance concerns in Chinese models.<\/p>\n
Any restrictions would be part of ongoing trade tensions between the U.S. and China that have extended into AI. Washington previously placed export restrictions on critical computing hardware and equipment to China before easing some limits, while Beijing is now focused on developing domestic technologies and urging Chinese companies to use them. The Trump administration has stated its intention for the U.S. to dominate the AI race.<\/p>\n
The administration cites cybersecurity concerns and is reportedly reviving efforts to restrict models from Chinese labs following the launch of open-weight models like Moonshot AI’s Kimi K3.<\/p>\n
Models like DeepSeek V4 and Kimi K3 are open-weight and offer low API pricing, with DeepSeek-V4-Pro charging $0.87 per million output tokens versus $50 for Anthropic’s Claude Fable 5. Self-hosting also keeps enterprise data in-house.<\/p>\n
Enforcement is difficult because the model weights are downloadable, mirrored across public repositories and torrents, and can be run offline in air-gapped data centers. Fine-tuning and distillation can further obscure a model’s origin.<\/p>\n