Trump weighs slow-motion ban on Chinese AI models via sanctions, liability rules

The Trump administration has spent recent months exploring indirect measures to restrict Chinese AI models in the U.S. market, according to reporting by Axios. Since 2025, the Department of Commerce, NSA, and White House have evaluated several options: placing Chinese AI labs on sanctions lists, issuing security warnings, and issuing executive orders that would impose security requirements and liability on U.S. companies that host Chinese models.

The groundwork for these restrictions began as early as summer 2025, when the Commerce Department drafted rules designed to protect domestic supply chains from Chinese open-source models. Advisers who favored a lighter regulatory approach initially blocked those efforts. However, two developments shifted momentum toward stricter measures: the release of China's Kimi K3 model and personnel changes in the White House. These events helped supporters of tighter restrictions regain influence over policy direction.

Crucially, the administration does not appear to be pursuing a direct, legally binding ban. Instead, a source close to the government told Axios that the approach is "slower and more durable," relying on procurement rules, sanctions threats, and public pressure campaigns to discourage U.S. companies from adopting Chinese models. This indirectness serves a strategic purpose: another source suggested the administration could focus regulatory pressure on potential cybersecurity backdoors and security flaws rather than banning the models outright. OpenAI strategist Dean W. Ball recently predicted this strategy, labeling it "FUD"—fear, uncertainty, and doubt. As Ball wrote, "You just create enough regulatory risk that every regulated enterprise backs off. You probably don't want to create so much regulatory risk that you scare off the hyperscalers from serving Chinese models; this will just drive startups to sketchier providers. There's a happy middle ground here."

Commercial interests underpin much of this push. U.S. companies are increasingly using Chinese open-source models because they are cheaper and nearly as capable as American alternatives. Restricting their use would protect the market dominance of Google, OpenAI, and Anthropic. The broader context amplifies the stakes: the AI sector is driving much of the U.S. stock market's gains under Trump, meaning that threats to the business of major U.S. providers could have significant market consequences. The article notes that open-source models do pose genuine cybersecurity risks, yet a U.S. ban would not eliminate those threats and would do little to curb them. Open models also support cyber defense, and Hugging Face reports they can outperform commercial models at that task—suggesting that restricting access could create risks of its own.

The Trump administration's emerging strategy reflects a pragmatic shift from direct prohibition toward regulatory friction. Rather than risk legal challenges or international backlash that a formal ban might invite, the approach uses security concerns and liability frameworks to create disincentives. A source quoted by Axios made the strategic calculus explicit: direct bans are blunt instruments, whereas layered procurement rules, sanctions threats, and public warnings allow for calibration—enough pressure to deter corporate adoption without pushing smaller players toward unlicensed alternatives.

The commercial stakes are substantial. U.S. companies are increasingly adopting Chinese open-source models because they are cheaper and perform nearly as well as proprietary American alternatives. Restrictions would protect the market position of Google, OpenAI, and Anthropic—companies whose valuations and stock performance have become central to the broader market rally under Trump. The body notes that the AI sector is driving much of the U.S. stock market's gains, creating a structural incentive for the administration to defend domestic providers against lower-cost competition.

However, the article also flags a tension: open-source models pose real cybersecurity risks, but they also support cyber defense work and can outperform commercial models at that task. Restricting access could create risks of its own. The strategy's success will hinge on whether indirect pressure—without formal prohibition—can shift corporate behavior while avoiding unintended consequences.