US Threatens to Sanction Chinese AI Models for IP Theft

What Happened?

US Treasury Secretary Scott Bessent stated that his country could impose sanctions on Chinese artificial intelligence models for alleged intellectual property theft. According to reports from TechCrunch and Gizmodo, Bessent promised concrete actions in the coming days or weeks, expanding the Trump administration's campaign to curb China's advances in artificial intelligence. Although specific mechanisms have not been detailed, sanctions could include adding Chinese models to the Entity List or applying tariffs on AI services. This threat comes amid growing concern in Washington over China's rapid progress in AI, especially after the release of models like DeepSeek-R1, which directly competes with Western systems such as OpenAI's GPT-4.

Why Is This Important?

This threat marks a significant escalation in the tech war between the US and China. Until now, sanctions focused on chips and hardware (such as restrictions on NVIDIA and ASML), but now they directly target AI models, including open-source ones widely adopted by the global community. If materialized, sanctions could disrupt the AI development ecosystem, affecting startups, researchers, and companies using models like those from DeepSeek, Alibaba (Qwen), or Baidu (ERNIE). According to Stanford HAI data, Chinese models account for approximately 30% of open-source models downloaded on Hugging Face, with DeepSeek and Qwen among the top 10 most popular. This means sanctions could have an immediate global impact, similar to the restrictions on Huawei in 2019, but in a sector even more integrated into the tech value chain.

Potential Consequences

• Fragmentation of the open-source ecosystem: Chinese models like Qwen or DeepSeek have been fundamental to democratizing AI. Sanctions could force developers to migrate to Western alternatives, reducing competition and innovation. For example, DeepSeek-R1 has been used by over 10,000 developers on platforms like Replicate and Modal, according to TechCrunch data. A forced migration to models like Meta's Llama 3 or Mistral could increase inference costs by up to 40%, according to estimates from analysis firm SemiAnalysis.
• Impact on global startups: Many startups rely on free or low-cost Chinese models. Sanctions would increase their costs and limit access to cutting-edge technology. For instance, startups in Africa and Latin America have adopted Chinese models for their efficiency in non-English languages. A University of Oxford study indicates that 60% of AI startups in Kenya use Chinese models. Sanctions could stifle innovation in these regions.
• Chinese retaliation: China is likely to respond with similar measures, affecting US tech companies operating in its market or relying on its supply chain. Apple, for example, generates 18% of its revenue in China (according to its 2025 annual report), and retaliation could include restrictions on iCloud services or iPhone sales. Additionally, China could restrict exports of rare earths, essential for chip manufacturing.
• Regulatory uncertainty: The threat creates a climate of uncertainty for investors and developers, who might delay projects or seek more stable jurisdictions. Global AI investments fell 12% in the first quarter of 2026 according to CB Insights, and this new threat could exacerbate the trend.

Historical Context

The US-China relationship in technology has been tense since the Trump administration, with sanctions on Huawei, ZTE, and more recently on semiconductor companies. The novelty here is the focus on intellectual property of AI models, an area where China has advanced rapidly. According to TechCrunch, the current administration believes Chinese models have benefited from data and algorithms protected by US copyright, though no concrete evidence has been presented. This move resembles the industrial espionage accusations against Huawei in 2019, but in the AI domain. Unlike chip sanctions, which mainly affect manufacturers, model sanctions directly impact developers and users, potentially creating a broader domino effect. Moreover, it recalls the 2018 trade war, when tariffs affected entire sectors, but now AI is the central battlefield.

What Should Readers Know?

For tech professionals, it is crucial to monitor upcoming Treasury announcements. If sanctions are imposed, they could include banning the export of Chinese models to the US, restricting companies that use them, or even measures against individuals. Companies integrating Chinese AI models should evaluate contingency plans, such as migrating to alternatives like Llama (Meta) or Mistral (France). Additionally, it is important to distinguish between open-source and proprietary models: sanctions could be selective. For example, DeepSeek-R1 is open-source, while Baidu's ERNIE is proprietary. Sanctions could also include prohibiting the download of model weights from repositories like Hugging Face, affecting thousands of projects. Investors should consider geopolitical risk in their AI portfolios. Finally, governments of other countries, such as the European Union, could be pressured to take sides, further fragmenting the global AI market.

"This is not just a trade war; it's a struggle for control over the future of artificial intelligence," commented an analyst from TheVortiq. "Sanctions on AI models could have a deeper impact than chip restrictions because they directly affect developers and end users."

In conclusion, the threat of sanctions represents a turning point in AI geopolitics. The coming days will be critical to determine whether this is a rhetorical warning or concrete action that will redefine the global tech landscape. The tech community must prepare for a scenario of fragmentation, where international collaboration in AI becomes severely limited.