US government examines measures against Chinese AI models

The US government is apparently considering measures against Chinese AI. The performance of the open AI model Kimi K3 has revived previous efforts to prevent the use of foreign open-weight models in the US. Axios reports this, citing people familiar with the matter.

As early as last year, the US Department of Commerce is said to have considered placing several Chinese AI labs on the Entity List. US companies would then only be able to access their technology with government approval. In addition, restrictions on Chinese open-weight models and government warnings about possible security risks, intended to deter US companies from using them, were discussed.

Representatives of a regulation-skeptical course had stopped the plans at the time. However, since their departure from the government and the emergence of powerful Chinese open-weight models such as Kimi K3 and Qwen 3.8, security policy hardliners have regained influence.

Open AI models become a political issue

Instead of a formal ban, the government is considering, among other things, stricter procurement rules, a possible addition of Chinese providers to the Entity List, and public pressure on US companies using their models, Axios reports. At the same time, it wants to highlight possible backdoors and security vulnerabilities more strongly. Such measures could significantly dampen the use of Chinese open-weight models and the business of US companies dependent on them, even without an explicit ban.

Chinese open-weight models are increasingly becoming an attractive alternative to domestic offerings for US companies. Their cost advantage and increased performance are increasing competitive pressure on American providers. For Washington, this harbors not only possible security risks but also economic disadvantages: US providers could lose market share, while American companies would become increasingly dependent on Chinese technology.

OpenAI strategist warns of “AI Communism”

According to Axios, leading American AI labs or actors close to them approach the US government with new proposals every three to five months, which amount to a ban on open AI models.

Over the weekend, OpenAI's Head of Strategic Futures, Dean Ball, joined the debate, apparently supporting regulatory measures against Chinese open-weight models. In his post on X, he warned of a possible “AI Communism”, where AI would become state-provided digital infrastructure rather than a market product.

David Sacks, a tech investor and AI advisor to the White House, suggested such warnings were an attempt at market protectionism. The leading providers of closed models – referring to OpenAI and Anthropic – already form a duopoly in AI model revenues and are now trying to eliminate their open-source competition with the help of the government, Sacks criticizes.

Technology analyst Ben Thompson, on the other hand, considers the economic concerns about Chinese open-weight models to be exaggerated. The problem, he writes, is that American developers are increasingly dependent on Chinese models due to a lack of competitive domestic alternatives, Thompson writes. Therefore, instead of restricting open model development, the US should strengthen domestic open-weight development.

(olb)