The Deepseek Shock to U.S. AI Security

On April 16, 2025, the House Select Committee’s report titled “DeepSeek Unmasked” presented a scathing investigation into the Chinese AI chatbot DeepSeek and portrayed it as a covert instrument of the Chinese Communist Party (CCP) aimed at undermining U.S. national security. [2] With evidence from cybersecurity researchers, industry insiders, and U.S. defense agencies, the report warns that DeepSeek’s development reflects a broader CCP strategy to accelerate military-relevant AI innovation while evading export controls. [3] The Committee alleges DeepSeek of siphoning American user data to servers in China, manipulating information to align with state narratives, and leveraging stolen U.S. technology and restricted Nvidia chips to build its AI capabilities. [4] In response, the Committee calls for urgent reforms to export enforcement, greater interagency coordination, and proactive measures to counter what it deems a growing threat from China’s integration of AI into its geopolitical ambitions. [5]

A Grand Debate for AI Infrastructure Policies

Silicon may be the circuitry, but ideas are the true battleground: as Washington and Beijing tout innovation and fortify strategic chokepoints, the race for AI supremacy has become a live-fire test of geopolitical theory itself. Is liberal interdependence still powerful enough to knit research labs, chip fabs, and cloud stacks into a mutually enriching commons—or will realist logic, with its zero-sum calculus of power, fracture the AI ecosystem into rival blocs? The stakes stretch far beyond algorithms: whichever narrative prevails will script the geopolitical operating system of the 21st century.

A Liberalist Agenda for AI Competition

Liberal theory emphasizes economic interdependence, international institutions, and transnational actors as key influences on global public policy. [6] In a world of complex interdependence, multiple channels connect societies from trade and investment to information flows; governments interact with domestic and international firms through supply chains, finance, and innovation, which fosters a mutual interest in stability and cooperation to protect shared prosperity​. [7]

In a global competition for AI, private technology and AI firms are seen as pivotal transnational actors. They drive innovation, connect markets, and even influence national security. Liberal theory posits that as these companies span borders; they create informal ties between governmental and non-governmental elites and knit countries together in webs of interdependence. [8] AI research communities and tech supply chains are inherently global, making outright decoupling costly for all sides. Public-private networks emerge as well, as governments partner with tech firms or consult them in policymaking. From a liberal perspective, such networks and the firms themselves can be forces for peace: they can lobby against disruptive trade wars, push for common standards, and prefer a predictable international environment conducive to business. [9]

From Google’s DeepMind and OpenAI in the U.S. to Baidu and Alibaba in China—AI companies are at the heart of the competition. These firms invest heavily in research and often collaborate with global talent; cutting-edge AI advances often originate in the private sector instead of government labs​. [10] Liberal theory views this as a validation of the importance of domestic actors: the state that best harnesses its private innovators can gain an edge in power. [11] The U.S. approach has been to let the market lead and then leverage those advances for national ends—a bottom-up model. [12] China, by contrast, has pursued a more state-driven yet still hybrid approach (sometimes called “military-civil fusion”), by encouraging private AI champions but closely steering them with industrial policy. [13]

Despite rivalry, there are also transnational linkages: Chinese tech companies until recently built AI research teams in Silicon Valley, and Western firms tap into China’s vast market for data and talent. [14] Liberal theory would highlight these interdependencies as buffers against total conflict—each side benefits from the other’s ecosystem to some degree. Both countries would realize the value of international AI standards and institutions: for example, forums like the Global Partnership on AI involve private tech leaders and academics alongside governments to set ethical guidelines and interoperability standards. [15] In essence, the liberal view sees AI firms as agents, who can either feed a zero-sum arms race or, if properly engaged, become bridges for dialogue and joint progress. 

Point of Divergence: National AI Competence

However, mounting geopolitical tensions see moves to compartmentalize AI spheres: export controls on advanced AI chips to China, and China’s push to replace foreign tech with domestic alternatives. [16] The rise of China and the ensuing U.S.–China great power competition has extended the arena of rivalry into the high-tech and corporate domain. AI firms become strategic assets that can augment a state’s power or, if aligned with an adversary, pose a potential threat.

The Meta Testimony and Deepseek

On April 9, 2025, in testimony before the U.S. Senate Judiciary Subcommittee on Crime and Counterterrorism, former Meta executive and whistleblower Sarah Wynn-Williams accused the social media giant of cooperating closely with the Chinese government to implement censorship on its platforms. [17] She further alleged that Meta leadership, including CEO Mark Zuckerberg, knowingly ignored concerns from engineers about potential access to American user data by Chinese officials. [18] She claimed Zuckerberg was deeply involved in cultivating Meta’s relationship with China, including learning Mandarin and regularly meeting with Chinese-speaking staff. [19] More importantly, she warned that Meta’s open-source AI model, Llama, may have assisted the development of DeepSeek. [20] 

According to Wynn-Williams, Meta executives deliberately built a physical pipeline between the U.S. and China despite warnings about potential risks to U.S. user data, and only Congressional intervention prevented Chinese access. [21] She claims that Meta’s internal efforts, dating back to 2015, included direct briefings to Chinese Communist Party officials to assist China’s technological advancement–which positioned Meta’s involvement as a clear factor in China’s recent AI military projects. [22]

Realist Theory and National Security

Realism begins with the assumption of international anarchy: there is no higher authority above states to enforce rules or ensure security. [23] In an anarchic world, states must rely on self-help, and they remain the most important actors; private tech companies, despite their wealth and global reach, operate within this state-centric system. [24] In the 21st century, technological prowess is a key source of national power, so states treat leading tech companies as part of their national power base. Indeed, today’s major tech firms rival states in influence–some are “geopolitical actors with more resources and power than most nation-states.” [25] However, realism holds that they will not be left unchecked: great powers will seek to harness the innovations and wealth of the private tech sector for national ends. [26]

Under realism, even purely defensive or commercial advances can be perceived as threats by rival states, due to the security dilemma. A security dilemma arises when actions taken by a state to increase its security–for example, developing a new technology or bolstering its tech industry–are interpreted as offensive or capability-enhancing by other states, which will be tempted to respond in kind. [27] In U.S.–China relations, this dynamic is palpable: “actions on one side make the other feel less secure and push it to develop countermeasures.”​ If one country invests heavily in next-generation networks, AI, or quantum computing to secure an edge, the other fears falling behind and accelerates its own efforts. 

Neither can risk trusting the other’s benign intent; there is always uncertainty about how a new technology might be used–today’s telecommunications firm could enable tomorrow’s military communications. Each state’s attempt to achieve security through tech superiority inadvertently makes the other less secure, which furthers mutual suspicions. [29] The result is a self-reinforcing competition: one side’s gains in advanced technologies drive the other to develop countermeasures. [30]

In realist framework, non-state actors like private tech companies are not independent drivers of international politics, but rather potential tools of state power. States remain the unitary actors who ultimately matter, and they will use whatever assets they can to enhance their relative position. [31] The rise of colossal technology firms in the U.S. and China therefore gets filtered through state interests. These corporations possess tremendous capital, talent, and intellectual property–resources that can significantly bolster a nation’s power base. [32]

In theoretical terms, realism provides a stark interpretation of U.S.–China rivalry in the high-tech arena. Anarchy forces each state to depend on self-help and contest every advantage; the security dilemma ensures that even innovative advances breed mistrust and counter-action; power maximization drives both states to seek unparalleled technological might. [33] Within this framework, private tech companies are not neutral players but are subsumed into the grand strategic competition as instruments of national power; their role is to be mobilized and guided by states in the pursuit of security and hegemony. [34] As the trajectory of this competition will significantly shape the global balance of power in the 21st century, whichever state best harnesses its private tech sector for strategic ends may well translate its technological edge into “surpassing state power,” thereby gaining the upper hand in the rivalry. [35]

AI Policies Transition Forward

The DeepSeek shock has brought into focus mounting tensions at the heart of AI infrastructure and geopolitics, where the liberal promise of global interdependence is steadily giving way to a realist calculus of power and control. As U.S. policymakers confront the risks posed by foreign access to critical technologies, the line between market-driven innovation and state-led technological mobilization is becoming increasingly blurred. This transition signals a broader realignment of policy priorities, in which the protection of national interests is coming to outweigh the earlier emphasis on open cooperation. Moving forward, we will turn to the first 100 days of U.S. AI laws and policies to trace how this shift is beginning to take institutional form to assess the contours of an emerging framework for AI governance in a more competitive geopolitical landscape.