Analysis

China has unveiled its plans for a world run on and by Chinese AI. Speaking at the World AI Forum in Shanghai, Xi Jinping emphasized that his priority is the rapid diffusion of Chinese open-source models to dominate the global AI stack over US competitors. Xi’s speech was important not because it unveiled new technologies, although those occurred alongside the forum, but because it demonstrated the CCP leadership’s commitment to AI as an integrated infrastructure that China is selling to the world. The message was clear: Xi is positioning China as a cooperative, open-source partner compared to the US’s closed and exclusive models. 

Yet this message was undermined by the concurrent founding of a China-led World Artificial Intelligence Cooperation Organization, which made clear who is included in this vision. The organization’s 29 signatories include Cuba, Russia, and Venezuela. Even amid calls for openness, the CCP’s intentions involve a world hooked on Chinese AI and which favors China’s partners. Moreover, Xi hinted in his speech that the CCP would take a dynamic approach to model restrictions if required, which points to a potential future tiered system whereby open models dominate, but more powerful models face similar restrictions to their US counterparts. The question then will be where and how the CCP decides to introduce restrictions or to prioritize certain approaches to AI over others. 

China’s efforts to dominate AI will be a feature US-China competition for the foreseeable future. Xi Jinping’s speech, far from rhetorical bluster, is a map for how Chinese AI will attempt to dominate global markets and undermine foreign businesses to China’s advantage. If companies across the world decide which models to use based mainly on cost, then China’s models will win, particularly in poorer countries where US models are cost prohibitive. While Kimi is currently behind US models in terms of speed, this cost dimension will likely play a major role in how US frontier labs roll out additional models.   

The challenge for US businesses is that companies across the world will likely have access to these cheap models just as US companies’ access becomes potentially more restricted. Based on US government statements, it appears likely that the White House will seek to ban Chinese open models on national security and economic security grounds. In part, this is based on accusations that recent models, including MoonshotAI’s newly released Kimi, were built through the distillation, which is effectively  IP theft from US models. That is grounds enough for the White House to block Chinese models in the United States, and will likely feature prominently in Treasury Secretary Scott Bessent’s planned talks with China about AI in September. US companies, while tempted to use cheaper Chinese models to secure business advantage in the short term, will therefore likely find that door shut to them in the coming weeks or months.

Analysis

This month, 2430 Group spoke to the Small Business Investment Company Critical Technology (SBICCT) Initiative in Washington, DC about how to better protect US businesses from PRC-based IP theft. Bringing together US government representatives with heads of US defense-related investment firms, the event is part of a pioneering initiative to align the private sector with the US government’s new legal restrictions on investing in PRC-owned or directed industries. While the focus of the meeting was on the evolving US regulatory environment, the risk to US business from engaging with the PRC is far greater than compliance; it is the loss of business-critical intellectual property, which in some cases means survival.

US investors and small businesses need to understand that government regulations for engaging with PRC-based (and other foreign) firms and research organizations are tightening, particularly for the defense and tech sectors. This poses risks that investors or companies contracting to the US government could fail to comply with shifting regulatory requirements. Firms that do not conduct adequate FOCI investigations into their partners, suppliers, or investment targets could lose or disqualify themselves from contracts. The current challenge  is that most restrictions are coming via executive orders and executive department regulatory additions and changes, rather than by law; and the executive branch moves much more quickly than congress. This means that US firms need to be aware of these rapid changes, specifically suppliers or other partners that are newly added to BIS entity, 1260H, or other government restriction lists. Keep in mind that sanctioned organizations often disguise their ownership through holding companies, name changes, and governance structures. Compliance remains an obligation, even if a US company is unaware of the risk. 

Compliance, however, is a small subset of the total risk of PRC engagement. US companies and investors in key sectors face an unrelenting onslaught of PRC state-supported and coordinated effort to put them out of business. US companies must therefore assess and mitigate PRC FOCI risk beyond compliance. For investors, the challenge is that such FOCI assessments can appear expensive or unhelpful when a deal otherwise looks attractive or faces a tight turnaround schedule. Moreover, complete decoupling from the PRC is often neither possible nor desirable. While minimizing exposure to PRC entities is advisable, the reality is that this is not always practical for some investors or US businesses that, at least for now, rely on the PRC for suppliers, technologies, or other necessary business components. Investors will increasingly need to balance this position against constant FOCI and regulatory risks that arise from PRC connections. This means that FOCI and other risk evaluations will be increasingly necessary to minimize these companies’ potential legal, financial, and IP exposure.

Analysis

Decisions about whether to hire foreign technical experts have been with the United States since its founding. On the one hand, the United States has benefitted tremendously from bringing foreigners to the country. Imagine US science and industry without Alexander Graham Bell, Nikola Tesla, Albert Einstein, Chien-Shiung Wu, and Sergey Brin, all of whom were born abroad. During the Cold War, the brain drain from the Soviet Union was a boon for the United States and its allies. Without that movement of scientists like Roald Sagdeev and Valentin Turchin, the current primacy of American industries, from semiconductors to aerospace technologies and biomedical sciences, would not have been possible. 

On the other hand, some governments, notably the PRC, exploit their citizens to acquire confidential commercial and scientific information acquired in the United States. While those risks are real, it is important to remember that the fraction of Chinese-born employees who take part in that effort is tiny. United States technology and science have long benefitted from the pipeline of foreign expertise, ability, and drive, and black-and-white thinking on these issues risks undermining the future of US competitiveness.

In the current Sino-US Cold War, however, many of the world’s most promising researchers are choosing China over the US, and US companies are losing critical expertise and therefore competitiveness. China flaunts its talent recruitment successes, both to build the narrative of its superiority to the US and advertise the perks of such opportunities to potential recruitment targets. For instance, former chair of the University of Colorado Department of Chemistry, Zhang Wei, resigned his position to move to China’s newly founded Great Bay University, a “new-type” research university focused on science and technology and backed by significant state investment. 

Former director at Google DeepMind and architect of AI systems at Google, Apple, and IBM, Cao Liangliang, also left the US to join Hong Kong Polytechnic University, which Chinese state media describes as “a homecoming for Hong Kong’s artificial intelligence community.” Former Open AI researcher Yao Shunyu was also poached by Tencent to become their Chief AI Scientist, bringing with him significant expertise and familiarity with US tech developments. It is not just Chinese nationals. Omar Yaghi, a Nobel-prize winning Chemist and former faculty member at the University of California, Berkeley, left the US for a position at China’s Tsinghua University after federal cuts for science funding. 

China wooed these stars via a combination of compensation, prestige, research funding, and family ties. Zhang was promoted from Chair of a department (a typically undesirable role due to its additional administrative duties) to a deputy dean and a coveted, prestigious endowed professorship; Cao was also promoted to an endowed position at Hong Kong PolyU; and Yao moved from researcher to Chief Scientist. Zhang also mentioned family motivations behind his move. In the case of Yaghi, Tsinghua cultivated him over several years, appointing him as an honorary professor in 2022 and funding his research before recruiting him.

US businesses and research institutions can slow and reverse the drain by taking the initiative to identify and keep their best people. This may seem obvious, but too often American organizations simply assume that the United States is the default preferred destination of top scientists and engineers. This is no longer a safe assumption. A recent Pew study found that many countries now view China more positively than the United States. 

The US government can play a role in this regard as well. The administration recently finalized a new immigration rule that would limit international students to four-year visas, meaning that students wishing to pursue five-year BA degrees, MAs after completing college, or a PhD (which typically takes more than four years) will no longer be able to complete their course of study. Such policies are intended in part to protect frontier tech knowledge, which is a worthy goal, but these policies could produce the unintended consequence of transferring knowledge from the United States to China, rather than the other way around. China, by contrast, created a fast-track for STEM-related visa applications last year.

US companies have been winning the competition for international scientific and technical expertise since the 1930s, and this expertise has been central to American economic primacy. The reversal of the brain drain jeopardizes the US position.

Analysis

MoonshotAI’s Kimi K3 release is China’s latest DeepSeek moment. In a supposedly similar display of efficiency and nimbleness winning over raw compute power, Kimi’s open model infrastructure is itself impressive in its capacity to compete with US models. Much like with DeepSeek, however, Kimi’s importance is less about an individual model and more about the demonstrable emergence of Chinese model infrastructure, with dozens of similar open model startups emerging alongside MoonshotAI. Kimi is proof that China is intent on developing systems that enable the rise of such startups en masse. These systems will continue to produce such startups regardless of whether their models result from domestic innovation or from distillation from US models, as the US government has accused MoonshotAI doing with its latest Kimi model.

One reason that Kimi is causing apprehension in the US tech community is that it highlights that Chinese frontier companies are closing the US’s estimated three-month advantage over foreign competitors. Chinese models are rapidly heading towards recursive self improvement (RSI), where AI systems can redesign or enhance their own capabilities largely independent of human input, a potentially revolutionary security challenge for companies and governments alike. The race for RSI will likely define the coming months of the AI race, but it will depend on more than one company or model. Instead, as Xi Jinping emphasized at the World AI Forum, Kimi will likely be one of multiple Chinese firms looking to achieve RSI or equivalent, and will almost certainly enjoy immense direct state subsidies, cheap energy, and facilitated distillation from the Chinese government to help achieve those ends. In other words, China’s advantage is no longer principally about distillation, but rather about its ability to combine distillation with computing power and homegrown data.

What does China matching the US in AI output mean for US businesses? For investors, US models will require immense capital investments to not only continue to innovate, but also to stay afloat as the American closed models try to compete in the market against far cheaper Chinese open models. In a sign of changing times, for example, the Chinese government formally announced “token economy” and “token factory” as part of its policy platform, with the ambition that Chinese AI tokens exploit immense state subsidies to become a cheap provider of low-cost, high-volume AI tokens both at home and abroad. Chinese open models are therefore likely to increasingly arise as a business tool among US companies and their competitors. For US companies, a calculation will likely come down to how much data they are willing to hand over to these AI models–and by extension the government of China–in an effort to stay ahead of their competitors. 

US frontier labs will need to consider what Kimi, and Chinese models more broadly mean for their business strategies. Competition for talent and capital will heat up. US labs will need to overcome compute constraints, particularly as public backlash against data center construction continues across the United States. If the United States hopes to maintain advantage against China in AI, it needs not just private investment but government financial support for R&D, and a more comprehensive approach to how to restrict Chinese models. How to do this remains debatable, but a failure to act now will fundamentally undermine US competitiveness and the ability for US companies to respond to the threat from Chinese competitors.

Analysis

The car you are driving collects, stores, and transmits information about who you are, where you live, where you go, how fast you drive, what you say, and even what you weigh. Most modern vehicles are equipped with blue-tooth connected microphones to enable hands-free operation of paired phone and infotainment systems. Many vehicles increasingly contain cabin-facing cameras that track your eyes and body position for evidence of intoxication and fatigue. Outward facing cameras capture the surrounding environment, including sensitive locations. The vehicle’s apps and paired apps also collect vast amounts of data, and 84 percent of car manufacturers sell that data to third parties.

This is as true for American, German, South Korean, and Japanese cars as it is for vehicles manufactured in China. However, a bipartisan group of US senators and representatives have introduced a bill, which recently passed committee, that would ban Chinese manufactured vehicles, or any vehicles with certain Chinese parts from sale or even operation in the United States. 

Why only Chinese vehicles? The reason for the distinction is that the PRC requires all PRC based companies, including its automobile manufacturers, to provide all requested information and assistance to the intelligence services and national security agencies of the PRC. Such information about US politicians, military, intelligence, law enforcement, and counter-intelligence personnel, as well as their facilities, has immense intelligence value. Furthermore, networked vehicles are not simply mobile collection, storage, and transmission platforms; they can also be used for military “communication, guidance, jamming, and ground transport disruption, especially when combined with Low Earth Orbit satellites (LEOs) and ground networks.” In the case of Chinese vehicles, this ban will be particularly challenging for companies like Mercedes Benz, which is 20 percent owned by PRC investors, and which will likely face restrictions on sales in the United States if this bill becomes law. 

Finally, while some observers distinguish economic from national security arguments for banning PRC vehicles in the United States (and other countries), the CCP recognizes no such distinction. The significant, direct and indirect subsidies provided from PRC provincial and central authorities to EV companies in the form of cash, land, capital, and depressed wages gives China-based EV manufacturers a tremendous price advantage over foreign competitors. This is part of a PRC strategy to put foreign competitors out of business, which in turn is part of China’s national security strategy. 

Congress is increasingly focused on the PRC’s blending of commerce and espionage, and so automobiles are likely not the end of Congressional efforts to limit the import of Chinese-produced technologies. Passage of this bill into law, which is by no means certain, will encourage similar efforts. It is worth noting that laws are much more difficult to repeal than regulations or executive orders, and thus passing this bill would endure into future administrations.The US-Soviet Cold War lasted for half a century. The Sino-American version is likely to last just as long. Whether or not this particular law passes in its current form, the restrictions it proposes will be part of the legal and regulatory environment in the United States for decades. Manufacturers outside the PRC need to consider what this means for supply chains, future costs, access to the US market, and then plan accordingly.