A senior Republican senator has urged the Trump administration to adopt a coordinated strategy against China's growing influence in artificial intelligence, warning that American companies are increasingly relying on Chinese AI models despite potential national security risks. The concern centers on open-weight models, which are freely available for anyone to download and use, making them attractive to startups and enterprises seeking cost-effective alternatives to proprietary American systems.

What Are Open-Weight Chinese AI Models and Why Do They Matter?

Open-weight models are artificial intelligence systems where the underlying code and parameters are publicly released, allowing developers worldwide to build applications on top of them. Chinese companies like DeepSeek, Alibaba (Qwen), and Moonshot AI (Kimi) have released increasingly capable open-weight models that compete directly with American offerings. These models have gained significant traction in the global market because they offer strong performance at lower costs than proprietary alternatives.

Senator Tom Cotton of Arkansas released a letter to Commerce Secretary Howard Lutnick on Thursday, July 31, 2026, detailing his concerns about this trend. He cited a report estimating that nearly 80 percent of US startups build applications on Chinese models, pointing to real-world examples including Airbnb's chatbot powered by Qwen and Cursor's coding agent based on Kimi.

How Are American Companies Currently Using Chinese AI Models?

The adoption of Chinese AI models has spread across multiple sectors and use cases. Cotton highlighted several concerning examples that demonstrate how deeply these models have penetrated the American technology ecosystem. The most alarming case involves defense contractors using Cursor's coding tools, which rely on Kimi, to build software for the Department of War, even though Chinese models generate lower quality and more vulnerable code when the user is American.

Beyond defense applications, Chinese models are being integrated into mainstream commercial products. Microsoft was reportedly considering DeepSeek as an alternative for its Copilot Cowork platform, suggesting that even major technology companies are evaluating Chinese AI options for enterprise use.

What Security Risks Does Cotton Identify?

Cotton's primary concern centers on potential security vulnerabilities embedded in open-weight Chinese models. He warned that backdoors may be hidden within these models, and auditing produces limited assurances of safety. This risk is particularly acute for government contractors and defense companies, where compromised code could have national security implications.

The senator emphasized that American companies, particularly those doing business with the US government, should not jeopardize national security for higher profits. He noted that the Department of War already bars agency and contractor use of DeepSeek, and other federal agencies have blocked access internally, but these restrictions remain limited in scope.

Steps to Reduce US Dependence on Chinese AI Models

Cotton proposed a comprehensive, multi-pronged approach to address the challenge. His recommendations include specific policy actions and strategic investments:

• Export Controls: Maintain strict export controls on advanced computing hardware and semiconductor manufacturing equipment to limit China's ability to develop frontier AI models.
• Expanded Restrictions: Extend existing restrictions on Chinese AI models beyond DeepSeek to include other major players like Qwen and Kimi, preventing their use by government contractors.
• Government-Wide Ban: Enact a government-wide ban on contractor use of additional Chinese models, similar to the Department of War's existing DeepSeek prohibition.
• Competitive American Ecosystem: Support development of an open AI ecosystem in the United States that can compete directly with Chinese offerings in both domestic and international markets.
• Model Diversity: Ensure America offers a range of AI models for different applications, from state-of-the-art frontier models for defense and critical infrastructure to cheaper, more efficient models for commercial and consumer workloads.

Cotton argued that such a move would demonstrate that the US government will not accept the risks of Chinese AI while creating demand for American alternatives. He stated that customers at home and abroad will prefer an American option, which is how the United States can take the lead in global AI diffusion.

Why Does This Matter for the Future of AI Competition?

Cotton's letter reflects a broader geopolitical concern about AI dominance. He emphasized that the United States must not only lead in developing frontier AI models but also ensure that its technology becomes the global standard. This requires ensuring that as many US firms, researchers, and institutions as possible leverage AI for productivity and scientific gains, while also ensuring that global customers adopt and build on American models where the US can set standards, norms, and dependencies.

"I write to you concerned about American companies increasingly relying on Chinese AI models. While the US is the leader in the most cutting edge, closed models, recent progress in Chinese AI models highlights alarming gaps," Cotton wrote in his letter to Commerce Secretary Howard Lutnick.

Senator Tom Cotton, Republican Senator from Arkansas

The senator concluded by commending the administration's current efforts and expressing confidence that prompt action could help America keep its frontier advantage while unleashing a diffusion strategy that ensures US AI dominance, both domestically and globally. His recommendations suggest that the Trump administration views AI competition with China as a strategic priority requiring coordinated government action across multiple agencies and policy levers.