Trump administration weighs bans on Chinese open-source AI models

The Trump administration is considering placing restrictions that could stop US companies from running cutting-edge Chinese open-source AI models, a step that would hand OpenAI and Anthropic a stronger hold over the domestic market, Axios reported Monday.

The trigger is said to be the arrival of Kimi K3 from the Chinese lab Moonshot AI. The model landed on Friday in line with the World Artificial Intelligence Conference in Shanghai, with its developers claiming it matched up to the performance of Anthropic’s Fable and OpenAI’s ChatGPT for a far lower price. American firms have been switching to Chinese AI models for this exact same reason, as they cost less and do basically the same work.

Trump administration’s efforts to slow AI shift

Axios reported that the Commerce Department looked at adding several Chinese AI labs to its Entity List last year, a move that would cut off US access for these AI labs without a license from the government.

At around the same time, the National Security Agency and the White House Office of the National Cyber Director issued an advisory warning companies away from Chinese labs, and the Trump admin considered an executive order that would let US firms host Chinese models only if they guaranteed security and accepted liability for any breach.

None of these plans actually came to fruition, though, as officials worried the rules would choke innovation and consequently shelved all considerations.

However, things seem to be moving in a different direction now. Sriram Krishnan, a White House adviser who opposed government intervention, has left his role, and those interested in “tightening” national security now carry more weight in the conversations.

Low-key restrictions instead of outright ban

Sources told Axios the government may not need a formal ban to get the result it wants. Those familiar with the conversations described the likely playbook to be the invoking of procurement rules, renewed Entity List threats, and public pressure aimed at US companies that use Chinese models.

It is also seen as an effort to bring out possible backdoors and security shortcomings in those systems while pushing US developers to build a stronger open-source alternative of their own.

David Sacks, an outside White House AI adviser, argued on X on Sunday that the campaign serves the top AI companies and eliminates competition. “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,” Sacks wrote on X. He has repeatedly warned against regulatory capture that would benefit the largest labs in the AI race.

A source told Axios that those labs, or their allies, pitch the administration on banning open-source models every three to five months.

Chinese AI numbers worry Washington

Chinese models now handle 46.4% of routed token traffic on OpenRouter, an aggregation platform that lets developers switch between systems, compared to 35.7% for US-built models. DeepSeek alone stood at at 17.6% as of July 2026.

A Hugging Face study published March 16, 2026 found Chinese open-source models had 41% of all open-source AI model downloads.

Reports also stated that the startup Lindy dropped Anthropic’s models for DeepSeek V4, while Airbnb and Siemens are testing systems from Alibaba and DeepSeek to cut costs. The Trump admin is also partly at fault for this shift, as the export bans announced by the US in early 2026 on domestic frontier models, including Anthropic’s Claude Mythos 5 and Fable 5, left gaps that foreign models filled.

China is said to be selling compute at almost 10 times cheaper than the US, giving its models a clear value pitch even if the quality lags. OpenAI’s most recent valuation was at $852 billion and Anthropic’s $965 billion, with both planning to go public this year with trillion-dollar targets in circulation.

Kimi K3’s Moonshot is reportedly eyeing an IPO within six months at a valuation closer to $30 billion. Cheaper Chinese models, and cheaper Chinese listings, could increase the pressure on the US labs that need public markets to fund their expansions.