[International Politics in the Age of AI] ① US AI Strategy and Prospects for Military Application

I. Introduction

The United States perceives artificial intelligence (AI) not merely as a scientific technology but as a strategic asset that will determine the nation's survival and the future international order. In particular, the second Trump administration regards AI as a non-negotiable core technology within the context of US-China strategic competition, given AI's comprehensive impact on national security, economic growth, technological innovation, international influence, and alliance policy. Especially in terms of national security, AI is a key technology that can influence the outcome of future warfare, particularly intelligentized warfare. It plays a crucial role in areas such as lethal autonomous weapons, intelligence analysis, reconnaissance and surveillance, and cyber warfare, and is essential for securing decision-making superiority on the battlefield.[1] It will function as a means to secure US superiority in terms of balance, deterrence, and strategic stability against major competitors like China. In economic terms, AI is recognized as a core element of America's future global economic leadership, not only boosting productivity and creating new industries but also serving as a key technology for maintaining US national competitiveness, such as by increasing investment and attracting top talent to the US.[2] In this regard, technological superiority and independence are crucial, and reducing dependence on China's supply chains for semiconductors and computing infrastructure is essential. Leading the direction of AI technology development by setting technological standards has become a core task.[3] Furthermore, the US expresses concern over the authoritarian use of AI, opposes the use of AI by authoritarian states like China for surveillance and social control, and participates in the process of establishing related global norms.[4] These comprehensive efforts ultimately influence US international influence and alliance policy. In particular, the US seeks to expand its influence by providing technological partnerships and development cooperation to the Global South, and aims to secure an advantage in global strategic competition through cooperation with NATO and East Asian allies. The cooperation in this context will aim to enhance collective defense based on AI and improve interoperability for it, and will specifically seek to secure military superiority through AI-based command, control, connectivity, and integration.[5]

Currently, the US maintains a relative advantage over China in areas such as semiconductors, cloud infrastructure, large language models, human resources, and private investment ecosystems. For example, the monopoly of equipment from NVIDIA and ASML, global data centers like AWS and Google Cloud, and AI models such as GPT and Claude have formed the basis of the US AI ecosystem. Additionally, open-source platforms like Pytorch and TensorFlow, along with an open research environment, are accelerating the pace of innovation. Meanwhile, the US AI strategy, which has been driven by private sector innovation and supported by the government, is shifting towards an increased government role under the second Trump administration. Underlying this change is the perception that the US's relative advantage over China is temporary and that the gap with China is gradually narrowing. Compared to China's military-civil fusion policy, the US's national capability mobilization policy is still closer to a market-centric strategy, but the nature of technological diplomacy is slowly converging. In terms of the military application of AI, the US is attempting technological and military integration with its allies as a means of balancing and deterring China. This could lead to a conflict between the need for allied cooperation to respond to technological sovereignty and geopolitical threats. This paper, in this context, compares the AI strategies of the Biden and second Trump administrations, examines the organizational changes within the US Department of Defense and its military application directions based on the current state of AI technology development. Finally, it draws implications for South Korea's technological diplomacy and AI strategy.

II. Comparison of Approaches by the Biden and Trump Administrations

The objectives and priorities of the US AI strategy have varied slightly across different administrations. For instance, the Obama administration's White House report titled "The Future of Artificial Intelligence" showed an early perception of AI, emphasizing its importance for economic and industrial innovation, but the level of securitization was not high.[6] However, it expressed concerns about the fairness and safety issues of algorithms that could arise during AI usage. In contrast, the first Trump administration began to treat AI as a key means to determine the outcome of national security and the US-China hegemonic competition. On February 11, 2019, it issued the Executive Order on Maintaining American Leadership in Artificial Intelligence, and based on the National Defense Authorization Act for Fiscal Year 2019, the National Security Commission on Artificial Intelligence (NSCAI), established under the act and led by figures such as Google's Eric Schmidt and former Under Secretary of Defense Robert Work, discussed the importance of AI from a national security perspective. As a result, the NSCAI published its final report in March 2021. Based on this report, the securitization of AI began in earnest, including discussions on AI-related security threats, future warfare concepts, and battlefield changes due to the emergence of autonomous weapon systems.

Subsequently, the Biden administration emphasized safety, values, and technological control in AI utilization while pressuring China, whereas the second Trump administration focused on managing technological competition with China on a market-centric basis with keywords such as innovation, transactions, and investment. For example, the Biden administration focused on technological control through strengthened semiconductor export restrictions. The comprehensive export control measures announced in October 2022, restricting China's access to semiconductors and equipment essential for AI development, known as "Export Control on Advanced Computing and Semiconductor manufacturing Items to the People’s Republic of China," exemplify this approach by the Biden administration.[7] These export control measures were intended to delay China's AI technology development and its military applications, but[8] they have also been criticized for encouraging China's technological self-reliance and leading to the proliferation of relatively inexpensive Chinese AI technologies to the Global South, resulting in a failure to capture the global market.[9]

<Table 1> Key Documents Related to AI Policies of the Biden and Second Trump Administrations

In contrast, the Trump second-term administration prioritized deregulation and accelerating innovation in its AI strategy. For example, it implemented policies to sell American AI technology to developed countries through NVIDIA, expanding US market dominance, and in return, the government would recoup a portion of the profits.[10] Such deregulation policies have been met with positive reception from American industries and are considered a factor in attracting foreign investment.[11] Competition is also emphasized. For instance, to counter China's open-source AI models, the administration encourages the development of open AI models within the US and adopts an approach to provide greater support for the domestic AI ecosystem. A prime example is the Stargate project, which aims to invest $500 billion in AI infrastructure in collaboration with OpenAI, Softbank, and Oracle, reflecting an intention to secure AI infrastructure superiority over China. In this context, Executive Order 14179, “Removing Barriers to American Leadership in Artificial Intelligence,” and “Winning the Race: AI Action Plan,” both promulgated in 2025, outlined three policy pillars: securing global AI leadership, accelerating AI technology innovation for this purpose, building American-style AI infrastructure, and leading international AI diplomacy and security. Specifically, the goal is to expand the export of the US AI full stack and standardize American AI, implying a potential tension between incentivizing neighboring countries to develop sovereign AI and cooperating with the US.

Meanwhile, the Biden administration emphasized establishing norms and governance within its AI strategy. Executive Order 14110, mentioned in [Table 1], aimed to strengthen ethical standards in AI utilization, establish criteria for preventing discrimination, and create oversight mechanisms.[12] In contrast, the Trump second-term administration has been passive regarding value and norm-based discussions, exhibiting a perception that excessive regulation could slow down innovation. From this perspective, the Trump second-term administration can be seen as adopting a techno-realist stance, believing that it is more efficient for private companies to be responsible for their own AI safety measures and that excessive federal regulation hinders technological innovation.[13]

It is necessary to examine the differences in the perception and strategy towards AI between the Biden and Trump second-term administrations, as well as the background of these differences. Firstly, the Biden administration opted for a strategy of first establishing governance to prevent the risks posed by AI. In contrast, the Trump second-term administration prioritizes technological innovation and infrastructure development based on the urgency of maintaining technological competitiveness. Consequently, Biden focused on securing the stability of AI systems utilized at all levels of government and the private sector through regulation, while the Trump second-term administration focused on mitigating or eliminating such regulations, viewing them as impediments to innovation. Regarding infrastructure development, both administrations recognize its importance. However, the Trump second-term administration emphasized speed and efficiency in areas such as data centers, power grids, and semiconductor manufacturing, whereas the Biden administration also considered aspects of regulation, clean energy, and fair competition. On the international front, the Biden administration concentrated on building diplomatic AI governance and rallying like-minded countries for this purpose, while the Trump second-term administration has led diplomacy focused on securing competitiveness through technological superiority, export promotion, and technology sharing among allies.

In summary, if the Biden administration's AI strategy is based on a perspective of securitization combined with alliances, norms, and ethics, the Trump second-term administration aims to build selective partnerships based on market preemption, cost, and efficiency rather than norms. Changes in the Trump second-term administration's strategy can be attributed to concerns within the US technology and industry sectors regarding the Biden administration's AI strategy, which focused on regulation and export controls, and the heightened sense of threat due to the emergence of China's DeepSeek-V3 model in 2024.[14] Of course, this perception can also be seen as intertwined with the Trump second-term administration's characteristic "America First" policy, or economic nationalism, and its desire to secure an advantage in the US-China technology competition.

The Trump second-term administration's approach to AI is evaluated as increasingly government-led compared to the past, in terms of mobilizing national capabilities. This raises concerns about whether it is following a development trajectory similar to China's military-civil fusion system. However, it is not yet comparable to China's level in terms of the scope and intensity of government intervention, and the formation of institutionalized technology development structures. Considering China's situation, where all stages and areas—including research funding allocation, infrastructure investment, centralized technology strategy development, inter-agency coordination mechanisms, cooperation with state-owned enterprises, and strong regulation and control—are under government control, the US still emphasizes free competition among technology development companies and a policy of maximum deregulation. There is also an assessment that it is returning to the traditional US free-market-centered strategy that emphasizes innovation and autonomy.[15] Furthermore, due to the characteristics of the US industrial ecosystem, it has been pointed out that while direct government investment and industrial coordination might increase short-term efficiency, it could lead to long-term dependency on government support and inefficient resource allocation, particularly weakening the diversity and competitiveness of the startup ecosystem.

If the US AI strategy evolves towards a national-centric approach, the lines between defense, industry, and civilian sectors may blur, potentially hindering cooperation between allies and private companies, and intensifying the zero-sum nature of global AI governance competition.[16] However, in terms of technological diplomacy, the US and China exhibit similar trajectories, for example, in their efforts to export full-stack solutions and spread norms to allies and developing countries by integrating AI strategies with foreign policy. This will place pressure on allies to choose between the US and China and, as mentioned earlier, is likely to increase incentives for allies to develop sovereign AI.

III. The Process of Organizing US Defense AI and Defense AI Strategy

Investment in AI technology development at the US Department of Defense level began during the Cold War, but initial results in the 1970s-80s were very poor. In 2014, the US proposed the Third Offset Strategy, defining AI, robotics, big data, and biotechnology as technologies that would lead future warfare. This recognition was reflected in the 2018 National Defense Strategy, solidifying AI as a core defense capability.

Specifically, the role and status of AI have gradually increased over three phases[17]: First, the Project Maven period (2017-2018) represents a very rudimentary stage of military AI utilization, focusing on the automation of drone imagery analysis. Project Maven was rapidly adopted for practical use, particularly in counter-IS operations in the Middle East. The second phase began after the establishment of the Joint Artificial Intelligence Center (JAIC, 2018-22), which, building on the success of Project Maven, focused on establishing the first AI hub within the Department of Defense, expanding AI utilization, data integration, and AI-related education and training. From 2022 to the present, the Chief Digital and Artificial Intelligence Office (CDAO) has been established. This integration resulted from merging the JAIC, the Defense Digital Service, and the Chief Data Office. This integrated system aims to consolidate redundant AI projects and function as a unified governance framework linking AI and data. Furthermore, it seeks to transition to a data-centric warfare system in conjunction with Joint All-Domain Command and Control (JADC2) and ultimately achieve AI-based battlefield superiority.

As a result of this defense AI organization, the main institutions can be summarized as follows:

<Table 2> Major Institutions Related to Defense AI in the United States

The military application of AI pursued by the US Department of Defense has primarily focused on technological deterrence, command and control integration, and, in conjunction with these, balancing China through cooperation with allies. In practice, the US has collaborated with allies and partners in the Indo-Pacific region to research, develop, commercialize, and deploy emerging technologies such as artificial intelligence, cyber, space, and quantum. The objective is to deter Chinese military actions in areas like the South China Sea and the Taiwan Strait, and to secure a technological advantage that enhances this deterrence. These objectives have been pursued on three main fronts: first, defense technology innovation among US and allied militaries; second, enhancing interoperability and information sharing of emerging technology systems among allies; and third, cooperation between allies in the defense industry.[18]

Furthermore, for technological deterrence to be effective, policies, institutions, and funding mechanisms for applying civilian-led innovation militarily must be in place. Entities such as the DIU within the US Department of Defense, or venture capital firms related to the US Central Intelligence Agency (CIA) like In-Q-Tel, have played such roles. It is also necessary to form a technology ecosystem between the US and its allies. Multilateral forums such as AUKUS Pillar 2, QUAD, or US-South Korea-Japan cooperation have been avenues for technological collaboration. For example, the QUAD has established a 'critical and emerging technology working group' to advance AI-related technological cooperation, focusing on topics like cybersecurity and technology standardization, thereby strengthening interoperability and strategic alignment. In fact, the US Department of Defense has set achieving AI interoperability with allies as a key objective by 2025, which is interpreted as integrating AI technologies into joint operations and cooperative frameworks.[19]

For this technological ecosystem to effectively serve as a deterrent and a balance against China, multidimensional efforts are required, including the development of common doctrines and strategies, command and control systems capable of real-time information sharing and command transmission, integrated weapon systems and platforms, and AI-driven data fusion. Furthermore, for efforts in all these areas to materialize, U.S. allies must agree to military and technological integration with the United States, reaching a stage where this is guaranteed institutionally and legally.

For example, Mission Partner Environment (MPE) is an integrated digital network environment that enables collaboration through real-time operational information sharing between U.S. forces and allies. It refers to a shared operational network for connecting disparate information security systems and networks when conducting multinational joint operations through organizations such as NATO and the ROK-U.S. Combined Forces Command. It entails integrating information access rights and security classification systems so that allied forces can view the same screen and make rapid decisions, aiming to enhance situational awareness, realize joint command and control, and ensure swift decision support. This system is formed through data interoperability, role-based access control, and the establishment of cloud-based collaboration platforms. If AI is integrated into this system, it will evolve into a 'real-time decision community' among allies. For instance, it will automatically classify and summarize multinational surveillance and reconnaissance data and operational reports fed into the MPE. The Indo-Pacific Command is in the process of establishing a system to integrate and analyze sensor information from allies such as Japan, Australia, and South Korea through an AI-enabled Data Fabric. Furthermore, when building AI-based predictive models, it is anticipated that a common threat perception will be formed in multinational operations, and response strategies will be proposed. Project Maven has already demonstrated the practical application of AI for target identification. In essence, an AI-based MPE has been initiated, but full integration remains distant. It currently operates in a human-in-the-loop form, supervised by humans, and can evolve to a higher level through cooperation with allies.

Discussions are also underway regarding the authentication systems for accessing this MPE system. For example, AI Passport refers to a certification system that guarantees the reliability, transparency, and interoperability of allied AI systems.[20]It can be seen as a certificate for verifying reliability in multinational operational environments such as MPE or Combined JADC2 (CJADC2). Obtaining such a certificate will be essential to ensure AI interoperability during joint operations.

This situation presents numerous challenges for South Korea and the ROK-U.S. alliance. For instance, if South Korea develops sovereign AI for its defense sector, obtaining the necessary certificates for combined operations will be crucial, and maintaining a balance between sovereign AI and alliance-based AI passports will be necessary. Even without considering combined operations, extensive prior consultations will be required in various domains such as AI-based extended deterrence, and AI-based cyber and space security, including data sharing between South Korea, its allies, and partner nations, and the associated information networks, security policies, and classification systems.

IV. Policy Implications

In the context of the U.S.-China strategic competition, AI technology development is progressing rapidly. Moreover, as a means of balancing against China, the United States is attempting military and technological integration with its allies based on AI. This will lead to the dilemma of how to maintain alliance solidarity while preserving autonomous decision-making capabilities in an unstable Indo-Pacific security environment.

From South Korea's perspective, facing the military threats from North Korea and China, cooperation with the United States and partner nations is necessary. However, in terms of autonomy, technological sovereignty and national control over core data and AI algorithms are also required. Hanwha Systems, for example, has recently announced its intention to develop sovereign AI technology to reduce external dependence in South Korea's defense sector, and is also pursuing the establishment of domestic data centers, the development of large language models, and the creation of a semiconductor-linked AI ecosystem. Furthermore, concerns exist regarding accidental escalation due to automated intervention when developing AI-based command and control systems and response mechanisms.

Beyond these concerns, the feasibility of AI technology development and subsequent cooperation with allies remains uncertain. The current defense R&D structure is characterized by fragmented data distribution, making integration difficult, or data is classified as highly confidential, hindering AI training and integration efforts. Fundamentally, a doctrine for AI warfare and a legal/ethical framework are also absent within the Ministry of Defense. In this context, attempting AI-based defense cooperation with allies prematurely could be a hasty move or could lead to the relinquishment of technological sovereignty.

Certainly, the U.S.-China strategic competition will continue long-term, and consequently, prioritization of emerging technologies such as AI will persist. Considering this strategic environment, continuous investment and development in South Korea's AI technology development and military application capabilities, as well as the development of leverage with the U.S., will be necessary. Particularly in preparation for military cooperation with a U.S. administration that favors a transactional approach toward allies, such as a potential second Trump administration, measures to protect technological sovereignty through the establishment of an AI ecosystem and capacity building within South Korea should also be explored.■

[1]The White House. 2024. "Memorandum on Advancing the United States’ Leadership in Artificial Intelligence; Harnessing Artificial Intelligence to Fulfill National Security Objectives; and Fostering the Safety, Security, and Trustworthiness of Artificial Intelligence." October 24; US DOD. 2023. "DOD Releases AI Adoption Strategy." November 2; National Security Commission on Artificial Intelligence. 2019. "Interim Report To Congress." November 4.

[2]Executive Order 13859. 2019. "Maintaining American Leadership in Artificial Intelligence." February 11. (https://www.federalregister.gov/documents/2019/02/14/2019-02544/maintaining-american-leadership-in-artificial-intelligence?utm_source=chatgpt.com ); Rasser, Martijn et al. 2019. "The American AI Century: A Blueprint for Action." CNAS Report. December 17; U.S. Senate. 2021. "The United States Innovation and Competition Act of 2021 (USICA)." (S. 1260); U.S. CHIPS and Science Act. 2022. August 9.

[3]Kennedy, Mark. 2025. "America’s AI Strategy: Playing Defense While China Plays to Win."Wilson Center Report. January 24.

[4]National Institute of Standard and Technology. n.d. "AI Risk Management Framework." (https://www.nist.gov/itl/ai-risk-management-framework)

[5]Curtis, Lisa et al. 2022. "Operationalizing the Quad." June 30; Reuters. 2024. "With eyes on China, Us and Japan vow New Security collaboration." April 11; Segal, Adam. 2025. "The Future of Technology Deterrence in the Indo-Pacific."Perry World House Report. May 19.

[6]Executive Office of the President, National Science and Technology Council, Committee on Technology. 2016. "Preparing for the Future of Artificial Intelligence." October.(https://obamawhitehouse.archives.gov/sites/default/files/whitehouse_files/microsites/ostp/NSTC/preparing_for_the_future_of_ai.pdf)

[7]US Department of Commerce, Bureau of Industry and Security (BIS). 2022. "Implementation of Additional Export Control: Certain Advanced Computing and Semiconductor Manufacturing Items: Supercomputer and Semiconductor End Use; Entity List Modification." October 13.(https://www.federalregister.gov/documents/2022/10/13/2022-21658/implementation-of-additional-export-controls-certain-advanced-computing-and-semiconductor)

[8]Webster, Graham. 2025. "Inside the Biden Administration’s gamble to Freeze China’s AI Future."WIRED. August 14.

[9]Kahl, Colin. 2025. "America is Winning the Race for Global AI Primacy- for now."Foreign Affairs. January 17.

[10]Tech Radar. 2025. "15% of Nvidia and AMD china chip sales to go to US government." August 11; Ramkumar, Amrith. 2025. "Trump’s Ai Strategy Against China Get Its First Big Test."Wall Street Journal. August 2.

[11]Washington Post. 2025. "Silicon Valley’s bet on Trump starts to pay off." July 24.

[12]Financial Times. 2024. "White House Science chief signals US-China Co-operation on AI Safety." January 25.

[13]AP News. 2025. "From Tech podcasts to policy: Trump’s new AI plan leans heavily on Silicon Vallye industry ideas." July 24; Sukma, Isti Marta. 2024. "Techno-Realism: Navigating New Challenges in the Contemporary Role of Technology in Politics." Security & Defense Quarterly 46. February.

[14]Ko Jun-sung. 2025. "The Great Shift in US AI Policy Amidst the Tech Hegemony War." 『Diplomacy』 154: pp. 114-138.

[15]Friedler, Sorelle et al. 2025. "What to make of the Trump Administration’s AI Action Plan."The Brookings. July 31.

[16]Rehman, Iskander et al. 2025. "Seeking Stability in the Competition For AI Advantage."RAND Commentary. March 13; Chavez, Pablo. 2025. "US AI Statecraft."CSET Commentary. October.

[17]Kahn, Lauren A. 2025. "Risky Incrementalism: Defense AI in the United States."In The Very Long Game of Defense AI Adoption, ed. Heiko Borchert, Torben Chutz, and Joseph Verbovszky. Switzerland: Palgrave.

[18]Segal, Adam. 2025. "The Future of Technology Deterrence in the Indo-Pacific." The Perry World House Report. May 19.

[19]Mahoney, Casey. 2022. "Shared Responsibility: Enacting Military AI Ethics in US Coalition."The National Interests. April 30.

[20]US Department of War. 2024. "Artificial Intelligence Defense Technical Review Explores Scalability and Federation." July 11.

Jeong Gu-yeon_Professor, Kangwon National University.

Lim Jae-hyun_EAI Researcher

    Inquiries: 02 2277 1683 (ext. 209) | jhim@eai.or.kr