The geopolitical and regulatory complexities behind U.S. efforts in the global advancement of AI

In March, the United States Commerce Department added more than 50 Chinese technology organizations to its export blacklist as it doubles down on curtailing China’s artificial intelligence (AI) and advanced computing capabilities. The blacklist bars American companies from supplying to those on the list without government permits. The companies were blacklisted for allegedly acting contrary to U.S. national security and foreign policy interests, further restricting China’s access to computing technology, which can process vast amounts of data at high speeds, as well as quantum technologies. In response, China’s foreign ministry said it strongly condemns the export restrictions, while urging the U.S. to “stop generalizing national security.” Importantly, the rise of Chinese AI startup, DeepSeek, has enhanced China’s adoption of open-source low-cost AI models, competing with leading U.S. companies that have higher-cost and proprietary models. Further, Chinese firms have gained access to U.S. strategic dual-use technologies via certain third parties, through loopholes that allow Chinese companies access to U.S. technologies, despite restrictions. U.S. officials note that they will continue to increase tracking operations aimed at smuggling advanced semiconductors made by Nvidia and Advanced Micro Devices. For the United States, there are two factors to consider in the global advancement of AI and other technology: one geopolitical and one regulatory. Seemingly, there is a contradiction between policy and industry: for example, the Trump Administration is being laisse faire regarding AI policy, but stricter on hardware, including semiconductor, regulation.

U.S. efforts to Counter Chinese AI Advancement

The Biden administration imposed numerous export controls against China, on everything from semiconductors to supercomputers under its “small yard, high fence” policy. More recently, Under Secretary of Commerce for Industry and Security said the agency was “sending a clear, resounding message” that the Trump administration will prevent U.S. technologies from the misuse of high performance computing, hypersonic missiles, military aircraft training, and UAVs (unmanned aerial vehicle) that threaten national security.

In addition, the Department of Commerce’s Bureau of Industry and Security (BIS) added 80 entities to the Entity List from China and other countries for activities contrary to U.S. national security and foreign policy. Part of BIS’s objectives aim to restrict China’s ability to acquire and develop high-performance and exascale computing capabilities, as well as quantum technologies, for military applications, and to impede China’s development of its hypersonic weapons program. Specifically, the Entity List includes: 12 entities—11 under the destination of China and one under the destination of Taiwan—for engaging in the development of advanced AI, supercomputers, and high-performance AI chips for China-based end-users with close ties to the country’s military-industrial complex; 27 Chinese entities for acquiring or attempting to acquire U.S.-origin items in support of China’s military modernization. These entities have demonstrable ties to activities of concern, including the development of hypersonic weapons and the design and modeling of vehicles in hypersonic flight; seven entities located in China for acquiring or attempting to acquire U.S.-origin items to advance China’s quantum technology capabilities; and two Chinese entities for selling products to parties on the Entity List, including Huawei and affiliated entity HiSilicon.

In addition, this week, a U.S. House panel concluded that the U.S. government should expand export controls and other tools to slow down the progress of Chinese AI companies like DeepSeek, while also preparing for a future where those efforts fail. The House Select Committee concluded that DeepSeek’s website and app “acts as a direct channel for foreign intelligence gathering on Americans’ private data” with its direct ties to China’s security and surveillance infrastructure and its unchecked data collection practices. Notably, the report included the extensive financial and technological resources that went into building DeepSeek’s R1 reasoning model, as well as its potential risks to U.S. economic and national security.

Early this year, DeepSeek’s R1 release impacted U.S. markets, as Chinese companies were further along in AI technology than initially assumed. In addition, the release highlighted that the capital and compute-intensive approach taken by U.S. AI companies in building their own commercial models may be unwarranted, given that R1 requires less processing power for similar or better performance. Further, rumors speculate that the R1 cost only $6 million to build and was allegedly built as the side project of a quantitative financial firm with little experience in the AI field. These claims have been heavily qualified, but the House report makes it clear that DeepSeek was not an inexperienced startup.

DeepSeek is officially owned by the much larger Ningbo Cheng’en Enterprise Management Consulting Partnership, but it is effectively controlled by Liang Wenfeng, who owns a majority stake. Wenfeng is the head of High-Flyer Quant, a Chinese quantitative trading firm that invested at least $420 million in initial funding into DeepSeek, while expanding access to the firm’s Firefly supercomputing infrastructure with 10,000 NVIDIA A100 GPUs. In addition, DeepSeek is part of the Hangzhou Chengxi Science and Technology Innovation Corridor, a Chinese state-sponsored economic development agency that supports China’s own Silicon Valley and has ties to other strategic state-linked hardware companies, such as Zhejiang Lab.

In its report, the House Committee found that DeepSeek is “designed to spy on Americans, steal technology, and subvert U.S. law,” including reportedly using advanced Nvidia chips that “should never have ended up in China’s hands.” The House committee held meetings with U.S. industry leaders, who argued that DeepSeek employees used aliases and international banking channels to purchase accounts for U.S. AI models and used that access to copy the work and create their own AI models. Interestingly, OpenAI and other U.S. companies are subject to numerous domestic lawsuits that allege they similarly appropriated massive amounts of protected intellectual property while training their own algorithms. As well as: 13 entities, including under the destinations of China, for contributing to unsafeguarded nuclear activities; two entities in Iran and China are added for attempting to procure U.S.-origin items for Iran’s defense industry and unmanned aerial vehicle programs; and ten entities under the destinations of China, South Africa, and the UAE due to their links to the Test Flying Academy of South Africa (TFASA) —a party added to the Entity List on June 12, 2023—and the training of Chinese military forces using Western and NATO sources.

Further, the report found that DeepSeek censors anti-CCP sentiments and collects and sends user data, including chat history, device details and typing cadence, via backend infrastructure to China Mobile, a telecom agency that the U.S. designated as a Chinese military company. Like other social media apps, DeepSeek has tracking tools and other methods for extracting granular data about users. The House report recommends that U.S. policymakers tighten export controls on semiconductor chips and other relevant AI technologies, expand funding and resourcing for the Commerce Department’s export controls office, and prohibit federal agencies from purchasing or using Chinese AI models. However, experts warn that the U.S. should prepare for when efforts fail to constrain international rivals like China from building high-quality AI systems, warning that China remains capable of rapid innovation in advanced technologies, despite U.S. efforts to stop them.

DeepSeek was reportedly developed using over 60,000 Nvidia chips, which may have been obtained in circumvention of U.S. export controls. Further, public records show Nvidia CEO Jensen Huang directed the company to design a modified chip to exploit regulatory loopholes after October 2023 restrictions. The Commerce Department said that Nvidia’s H20 graphics processing units — designed to comply with previous U.S. restrictions — would now require export licenses. Nvidia says it has already halted exports of the GPUs, resulting in a quarterly charge of approximately $5.5 billion. In response to the report’s findings, the Committee sent a letter to Nvidia demanding answers about sales to China and Southeast Asia to examine whether and how its chips power DeepSeek’s AI models, despite U.S. export restrictions.

In addition, the report found that over 85 percent of DeepSeek’s responses are manipulated to suppress content related to democracy, Taiwan, Hong Kong, and human rights—without disclosure to users. The platform funnels American user data through unsecured networks to China, serving as a high-value open-source intelligence asset for China, and its infrastructure is linked to Chinese state-affiliated firms including ByteDance, Baidu, Tencent, and China Mobile—entities known for censorship, surveillance, and data harvesting.

As the U.S. expands control on Nvidia’s sales to China, Chinese domestic artificial intelligence chipmakers, like Huawei, could benefit. Analysts have pointed to Huawei as a leader in China’s race to find a Nvidia competitor. The U.S.-blacklisted company has been working on its own “Ascend 910” GPU series. Although Huawei remains “a generation behind in chips,” the company is “making waves” with the hardware that uses them. However, experts note that export controls have also hindered China’s ability to produce advanced GPUs at the same scale that Nvidia can through its partner Taiwan Semiconductor Manufacturing Company (SMC), the world’s largest contract chip maker. SMC’s chipmaking equipment includes U.S. technology, so thus far the company has complied with U.S. trade restrictions on Huawei and the shipment of advanced chips to China. That has left Chinese companies increasingly reliant on domestic foundries like SMC. However, SMC is under its own export controls, which prevents it from accessing some of the most advanced chipmaking equipment. Given those conditions, it is unclear if Chinese chip foundries would be able to supply enough H20 GPU alternatives to meet the demand of Chinese tech companies any time soon.

According to Counterpoint, it remains unclear how long different firms’ existing stockpiles will last, but they should provide Chinese chipmakers time to scale up their GPU manufacturing. Meanwhile, Huawei’s Ascend chip shows how China’s export controls have failed to stop firms like Huawei from accessing critical foreign tools and sub-components needed for advanced GPUs. Given reports of export control workarounds and Huawei’s GPU progress, semiconductor experts have questioned whether the restrictions on Nvidia will achieve their intended purpose. “U.S. controls on GPUs and semiconductor manufacturing equipment have primarily damaged U.S. companies, including Nvidia, while having marginal impact on the ability of Chinese companies to develop frontier AI models.” In addition, the controls have encouraged the Chinese semiconductor industry to become more innovative while “designing out U.S. technology.” Next month, Nvidia may face additional restrictions on exports under “AI diffusion rules,” first proposed by the Biden Administration.

The geopolitical complexity is also demonstrated through U.S. relations with other countries. Even as the Department of Commerce increased export restrictions, the Trump Administration is open to Emirati dollars to build out U.S. AI infrastructure, https://www.fdd.org/analysis/op_eds/2025/04/10/to-beat-china-america-must-balance-ai-innovation-with-national-security/despite close relations between Emirati and Chinese technology firms. On March 20, UAE National Security Advisor Sheikh Tahnoon bin Zayed committed to a decade-long, $1.4 trillion investment framework to fund next-generation data centers, semiconductor manufacturing, and energy infrastructure. For example, Abu Dhabi-based investment group MGX welcomed NVIDIA and xAI to its collaboration with Microsoft, BlackRock, and Global Infrastructure Partners as part of the AI Infrastructure Partnership, aimed to support global AI supply chains through joint financing.

Conclusion

U.S. export restrictions on NVIDIA prevent the company from selling H20 AI accelerators to China, and there are claims that NVIDIA's revenue to China is higher than estimated figures, given that trade with China involves several other nations, such as Singapore and Malaysia. Although rumors also indicate that NVIDIA plans to develop custom AI chips in partnership with DeepSeek, including HBM, process node, and packaging facilities, experts find these claims unlikely. Due to the geopolitical impacts of U.S. relations with Tawain and China, it is unlikely that Taiwan or the U.S. would engage strategically in this partnership. Further, Deepseek could be using NVIDIA hardware, but given how concerned the U.S. is over supply chains and semiconductors, it is unlikely.

However, the Trump Administration appears indifferent to security threats facing Taiwan; instead, it has expressed greater concern that Taiwan “stole” the semiconductor industry from the United States. U.S. calls for tariffs on Taiwan’s chip exports reportedly contributed to TSMC’s announcement of a $100 billion investment plan for new semiconductor manufacturing facilities in Arizona. In response to a question about whether TSMC’s Arizona investments would lessen the impact of a Chinese invasion of Taiwan, Trump said that “it will at least give us a position where we have – in this very, very important business, we would have a very big part of it in the United States,” suggesting that his resolve to defend Taiwan could be contingent on the onshoring of advanced chip manufacturing.

On the one hand, AI global development is a geopolitical issue. The U.S. wants to bring semiconductors to the U.S., allowing it to not have to rely on Taiwan. However, with proximity to Taiwan, China has access to the Pacific, and then the U.S. loses dominance in the region. No country can stop China from buying hardware from Nvidia, but it certainly would not benefit by making the U.S. angry. On the other hand, AI is a regulatory issue. The U.S. wants to keep ahead of AI development race, meaning they may be lax in AI regulations regarding ethical and other limitations market limitations. This could give China more opportunities, but it remains unclear