[International Politics in the Age of AI] ⑦ Prospects for the Development of AI Military Innovation: Two Perspectives on the Pace of Innovation and Cases from the US and China

⑩ AI and International Political Economy, Song Ji-yeon [Read Working Paper]

I. Introduction

Artificial intelligence (AI) technology is rapidly developing and spreading. This technology is expected to transform almost all areas of human activity, and the military domain is no exception. In particular, technological changes at the level of the Industrial Revolution have invariably led to revolutionary changes in the overall conduct of warfare.[1] This has triggered what is known as a Revolution in Military Affairs (RMA).

However, there are two opposing views regarding the pace and nature of AI-driven military innovation. On one hand, it is argued that due to the general-purpose enabling nature and innovativeness of AI technology, it will rapidly and innovatively transform not only weapon systems but almost all aspects of military operations.[2] Conversely, on the other hand, it is argued that, as seen in most past military innovation cases, it will take a considerable amount of time to fully realize the technological potential, and the pace of development may be gradual and uneven.[3]

Predicting the pace and pattern of future AI military innovation is crucial, both theoretically and practically. This is because it is essential to forecast the landscape of future battlefields for military planning and force development policy decisions that require a horizon of at least 5 years, and ideally 10 to 15 years or more. If AI-based military innovation progresses rapidly, rendering existing operational methods inefficient, continued investment in traditional platforms will incur opportunity costs. Conversely, if the transformation occurs gradually, a phased transition based on existing capabilities may be strategically rational.

Accordingly, this study will first outline the contrasting theoretical perspectives on the pace of innovation and then analyze the military innovation and defense reform trends in the United States and China based on these perspectives. This endeavor requires access to extensive research and classified materials, and thus has inherent limitations given the confined scope and nature of this study. Nevertheless, this study aims to provide meaningful insights and policy implications for future force development and strategic planning based on theoretical arguments and publicly available data accessible to date.

II. Theoretical Discussion: Two Perspectives on AI Military Innovation

1. Radical Innovation Theory: Rapid Transformation of Warfare by AI

The radical innovation theory fundamentally posits that artificial intelligence (AI) has the potential to fundamentally alter the nature of warfare. From this perspective, AI is not a single weapon like the longbow or tank, but is considered a general-purpose enabling technology comparable to electricity or the internal combustion engine.[4] It predicts that AI will penetrate various military domains, potentially reshaping weapon systems and doctrine as a whole. The emergence of AI is seen as a new driver of innovation that will drastically change the character and pace of warfare itself.[5] Indeed, both the United States and China are competitively accelerating the development of AI military technology, recognizing AI as a strategic technology that will change the future conduct of war. Particularly due to the generality and innovativeness of AI, existing weapon systems and operational concepts can quickly become obsolete. Proponents of radical innovation emphasize that the integration of AI into military power must lead to a reorganization of overall operational concepts, not just incremental performance improvements. Militaries that limit the application of AI to partial improvements of existing methods will inevitably fall behind those pursuing fundamental innovation, and decisive action to radically alter organizational structures and doctrines is necessary to gain an advantage on future battlefields.[6]

Furthermore, the pace of application of such technological innovation is being accelerated by the geopolitical environment of US-China strategic competition. The Chinese People's Liberation Army is making substantial investments to complete military modernization by 2035 and realize 'intelligentized warfare' centered on AI, accelerating the development of next-generation AI capabilities such as autonomous unmanned weapons, intelligent sensors, and networked command and control.[7] The rapid efforts by the Chinese military to operationalize AI, coupled with its aggressive posture in areas like the Taiwan Strait and the South China Sea, are having significant repercussions on the regional security environment. In response to China's rise, the US Department of Defense has also designated AI as a core future warfare technology and prioritizes rapidly introducing civilian innovation capabilities into the military domain. Amidst the competitive dynamic between the two nations, the race for AI military technology development is accelerating, and the question of who will first seize the initiative in future warfare through AI has emerged as a key issue in the US-China hegemonic competition.

Moreover, recent regional conflicts are acting as catalysts for the practical validation and proliferation of AI military innovation. Russia's invasion of Ukraine is being called the first war where new AI technologies were extensively applied by both Ukrainian and Russian forces, and it is evaluated as a case where AI-based drone capabilities and autonomous weapon systems were deployed on the battlefield ahead of international norms or regulations.[8] The real-time sharing of how even some low-level AI technologies can rapidly change the nature of warfare globally, as seen in the improvised use of tools by both sides in the Ukraine war evolving into state-of-the-art systems, is noteworthy. For instance, the Ukrainian military conducted maneuver warfare using autonomous drones that perform target identification with AI, and the Russian military also gradually increased its use of drone swarm tactics and AI-guided munitions in response.

It is also conceivable that these new tactics and technologies, once proven effective in the crucible of the Ukraine conflict, will be rapidly disseminated to conflicts in other regions. In other words, the urgent security situation in regional conflicts is forcing the early operational deployment of AI innovative technologies that might otherwise remain in the planning stages during peacetime, and their success stories are being rapidly emulated by militaries worldwide. Consequently, the radical innovation theory posits that AI military innovation is becoming an inevitable and irreversible trend, driven by the convergence of technological innovation and geopolitical pressures. The rapid advancement of AI technology, combined with the security dilemma of hegemonic competition, has created an environment where neither side can afford to back down from the AI adoption race, and the military application of AI is perceived not as an option but as a prerequisite for survival.

2. Gradual Innovation Theory: Delays and Gradual Changes in AI Innovation

The gradual innovation theory acknowledges artificial intelligence (AI) as a significant enabler of military power but suggests that it is more likely to diffuse in a gradual and uneven manner, constrained by institutional, organizational, doctrinal, personnel, and acquisition systems, rather than directly leading to a radical transformation that changes the 'foundation' of warfare in a short period.[9] In other words, even if the technology itself has great potential, its full realization into a structural transformation of warfare requires the military to undertake extensive learning, experimentation, organizational redesign, and force structure adjustments over a long period.

First, military innovation is not established merely by 'possessing' technology; it requires a concurrent reorganization of the operating organization, doctrine, operational concepts, personnel composition, training, and evaluation systems.[10] Specifically, AI relies on complex infrastructure including software, data pipelines, networks, security, authentication, and testing and evaluation. Therefore, rather than linear diffusion through "introduction-deployment" as with existing weapon systems, its performance and reliability are significantly influenced by data accessibility, quality, update cycles, and operational environment variability. These conditions are difficult to meet quickly within the inertia of peacetime bureaucracy and acquisition systems, and consequently, the pace of innovation may lag behind technological advancement.[11]

Second, the asymmetric advantage offered by new technologies can be quickly eroded. AI, in particular, is a general-purpose foundational technology, meaning it is likely to rapidly enable not only offensive advantages but also countermeasures (defensive) techniques and systems to neutralize or offset them. Despite the surge in drone operations on the Ukrainian battlefield, countermeasures such as electronic warfare (EW), jamming, and spoofing have rapidly intensified. Consequently, many unmanned aerial vehicles have lost their targets or failed their missions. In response, a competitive escalation of "counter-response" innovations, such as fiber-optic cable-based operations and AI navigation/targeting functions in communication-loss scenarios, has been observed at the battlefield level.[12] This mutual adaptation reinforces the core argument of incremental innovation theory: "the decisive advantage of new technology is difficult to sustain for extended periods."

Third, initial technologies are generally immature, have limited reliability, and cannot immediately replace existing weapon systems. While AI can exhibit strengths in specific tasks, in battlefield environments, performance can be rapidly degraded by changes in data distribution, enemy deception/disruption, sensor/communication constraints, and a lack of actual combat data. Furthermore, even if AI demonstrates strengths in tactical optimization, it is still far from automatically resolving the uncertainties of war and strategic contextual judgment. Consequently, in the short term, "augmentation of existing weapon systems" is more prevalent than "replacement of existing weapon systems," and the fundamental conduct of warfare is likely to transition not at once, but through a process of "partial automation – selective adoption – incremental adjustment of operational doctrines."

Fourth, from the perspective of defense budgets and acquisition systems, there are strong incentives to maintain the choice of "improvement" over "replacement" for extended periods. Large-scale weapon systems require significant time for planning, development, testing, production, and operational deployment. This long-cycle structure encourages continued investment in existing capabilities (upgrades, life extensions, improvements) even as new technologies emerge. Moreover, the longer the acquisition process, the greater the risk that the technology will become obsolete by the time it is deployed, making "how to incrementally integrate" rather than "what to replace" the focus of practical decision-making.[13]

Fifth, as warfare involves high risks, there are structural constraints on the full adoption of unproven methods. Incremental innovation theory posits that militaries tend to favor experimentation, verification, and doctrinal development within limited scopes before widespread adoption, rather than rashly implementing new concepts in areas with high "costs of failure." Indeed, despite the rapid proliferation of unmanned and automated technologies on the Ukrainian battlefield, the large-scale loss of unmanned vehicles under the pressure of electronic warfare and air defense has led to a tendency for unmanned systems to be associated with "high-loss, high-turnover, attritional operations" rather than "full replacement."[14]

Sixth, the reality of the war in Ukraine demonstrates that the "speed at which innovation changes the battlefield" and the "speed at which innovation changes the structure of warfare" can differ. At the tactical level, drones, sensors, and AI-assisted functions are altering battlefield visibility and strike procedures. However, simultaneously, the proliferation of robust defensive and disruptive measures, high attrition rates, and the importance of human control, judgment, and logistical systems remain central to warfare. Notably, while the "technological illusion" centered on unmanned capabilities may be amplified during peacetime, under conditions of technological parity and mutual proliferation, the emphasis on human, organizational, and industrial mobilization may re-emerge.

In summary, incremental innovation theory suggests that while AI will permeate military domains extensively, its effects will be constrained by (1) the pace of change in organizations, doctrines, personnel, and acquisition systems; (2) the rapid emergence of countermeasures and erosion of advantages; (3) limitations in data, reliability, and verification; (4) long-cycle force structures and budget constraints; and (5) risk aversion in warfare. Therefore, in the short term (around the next five years), "selective integration to enhance the performance and efficiency of existing forces" is likely to prevail over "a complete transformation of warfare by AI." In the medium term (10-15 years), innovation is likely to unfold asymmetrically and incrementally, at varying speeds across different nations, military branches, and mission areas.

III. Case Study: Current Status of Military AI Application in the United States and China

Based on the theoretical perspectives reviewed above, this chapter aims to analyze the current status of military AI application in the United States and China. As elaborated, this task inherently demands extensive research. Defining the substance and scope of military AI innovation is challenging even at this juncture.[15] Considering these limitations, this study seeks to synthesize recent research and evaluation findings on major policy initiatives related to artificial intelligence, automation, drones, and robotics in both the U.S. and China over the past decade.

1. Assessment of AI Military Innovation Status in the United States

In the mid-2010s, as competition with China intensified, the United States announced the "Defense Innovation Initiative (DII)" based on the concept of the "Third Offset Strategy," which aimed to create new qualitative advantages through advanced technologies, thereby accelerating the application of new technologies from the Fourth Industrial Revolution to the military domain.[16] The core of the innovation plan, announced by then-Deputy Secretary of Defense Robert Work in 2015, was "automation" and "robotics." Over the past decade, the United States has strived to adopt AI-based automation and robotics through various policy initiatives at the Department of Defense and Joint Staff levels, as well as within each military branch. This study will evaluate the achievements and limitations to date, focusing on Project Maven and Joint All-Domain Command and Control (JADC2).

Project Maven, initiated by the U.S. Department of Defense in 2017 as part of the "Algorithmic Warfare Cross-Function Team (AWCFT)" project, is a pilot program aimed at enhancing the speed and accuracy of military intelligence analysis by incorporating AI machine learning into the processing of information, such as drone imagery. Project Maven was initially deployed for the mission of identifying enemy targets in drone reconnaissance imagery during counter-terrorism operations in the Middle East, operating in a "Human in the loop" capacity to assist human analysts.[17] Despite setbacks, such as Google's withdrawal in 2018, it has since evolved with the participation of private companies like Palantir.

In terms of operational organization, Project Maven was integrated into the Department of Defense's Joint Artificial Intelligence Center (JAIC) in 2022, elevated to an official program under the National Geospatial-Intelligence Agency (NGA), and secured stable funding starting in Fiscal Year 2024. The Director of the U.S. NGA stated in 2023, after Maven's transfer, that "significant technological leaps have been made in a short period, and it is already contributing to America's most critical operations."[18] Indeed, in February 2024, the U.S. Central Command officially confirmed that "Maven AI algorithms contributed to the selection of targets for over 85 airstrikes conducted in the Middle East region during that month." It was revealed that AI was used to identify targets for airstrikes in Iraq and Syria, detect enemy rocket launchers in Yemen, and identify suspicious objects on the water in the Red Sea, leading to subsequent strikes.[19]

As demonstrated, Maven has shown tangible results in improving the speed of intelligence analysis in real-world operations. Furthermore, with NATO signing a contract to adopt the Maven-based AI battlefield system in 2025, international expansion cases are emerging, and[20] as successful cases accumulate, its scope of application is expected to broaden further. In conclusion, Project Maven has succeeded in technological maturity and organizational absorption, with some components already operationalized as U.S. military intelligence assets, and it is demonstrating significant achievements in actual combat.

The JADC2 initiative, pursued by the U.S. Department of Defense, is a next-generation command and control concept aimed at supporting rapid military decision-making through AI and data fusion by connecting all sensors to all firing platforms within a single network. In 2022, the Department of Defense established the JADC2 strategy and implementation plan, and in early 2024, it was announced that a Minimum Viable Capability had been achieved and deployed to U.S. Central Command (CENTCOM).[21] Indeed, in 2023, an early version of JADC2 was utilized in operations responding to drone swarms and missile threats in Yemen and Syria, reportedly contributing to real-time target engagement decisions by ships and fighter jets through a "Common Operational Picture" integrating information from multiple sensors.

While the JADC2 concept has entered the trial operation phase in some theaters, its full integration implementation is still considered to face challenges.[22] In its 2025 report, the U.S. Government Accountability Office (GAO) pointed out "the lack of clear category definitions and performance measurement systems for JADC2 implementation," noting that systems independently invested in by each military branch are not being comprehensively managed to align with the JADC2 vision. According to the GAO, the initial concept of "connecting all sensors to all fires" has been adjusted as unrealistic, and the lack of investment guidelines and performance metrics at the DOD level makes it difficult to objectively assess the progress of JADC2 implementation. In summary, JADC2 has begun to be utilized experimentally in some operational environments, but it has not yet reached the level of full-scale operational integration across the entire military.

The Replicator initiative, announced by U.S. Deputy Secretary of Defense Kathleen Hicks in 2023, is an innovation concept and implementation plan aimed at securing deterrence against China by dramatically increasing the number of autonomous unmanned systems within the next two years. This initiative is based on the concept of mass-deploying low-cost, attritable drones across multiple domains—air, sea, and land—as if "replicating" them, to overwhelm Chinese defenses with saturation attacks or paralyze their operational capabilities in case of conflict.

At the time of its announcement, the Replicator plan set an ambitious goal of achieving initial results by 2025 through the reallocation of existing budgets and significant streamlining of acquisition processes, without additional funding. In terms of innovation implementation, by mid-2024, the U.S. Department of Defense had selected several platforms for the first phase of Replicator, including the loitering munition "Switchblade 600" (Air Force), the small "Ghost-X drone" (Army), and "unmanned surface vessels," and announced their initial deployment to the Indo-Pacific Command.[23] By May 2024, some Replicator drones had already been delivered to U.S. forces in the Indo-Pacific region, and Deputy Secretary Hicks commented that "battle-centric innovation is yielding visible results."[24]

However, the success of its operational deployment remains uncertain. A 2025 report from the Congressional Research Service (CRS) cited an assessment by a former defense official suggesting delays compared to the initial target, stating that "the plan to deploy thousands of unmanned systems by the summer of 2025 has resulted in the deployment of only a few hundred by that time."[25] Furthermore, the Replicator initiative has faced challenges related to limited public disclosure due to the need for secrecy, hindering congressional oversight, and concerns about budget competition with existing programs. Technically, challenges remain in areas such as integrated command of multi-domain drone swarms, AI autonomy, and communication network security. Experts point out that, considering the bureaucratic hurdles within the U.S. defense acquisition system, achieving the objectives is not entirely optimistic.[26] Nevertheless, the Replicator initiative is evaluated as seeking change by impacting the U.S. military organization through a "pathfinder" approach that emphasizes extremely short-term results, unlike previous projects.[27] If successful in actual deployment, even on a small scale, within the next one to two years, it could serve as a catalyst for innovation. However, at present, it is appropriate to evaluate it as being in a partial testing phase.

2. China's Military AI Application and Drone/Unmanned System Operations

Under Xi Jinping's leadership, the People's Liberation Army (PLA) has designated "Intelligentized Warfare (智能化战争)" as a national priority, driving military innovation centered on AI. This concept aims to secure dominance in future warfare by leveraging advanced technologies such as AI, big data, and autonomous unmanned systems across all aspects of military operations, moving beyond mere informatization.[28] Specifically, the PLA has identified AI-based command and control systems, drone swarm tactics, and autonomous unmanned platforms as key areas for intelligent weapon development, dedicating the past decade to research, development, and field testing. The following section evaluates the technological progress and operationalization level of these representative Chinese innovation efforts based on assessments from within the United States.

The concept of intelligentized warfare within the Chinese military has been a core tenet of the PLA's modernization efforts since its appearance in the Central Military Commission's work report in 2017 and the National Defense White Paper in 2019. Intelligentization is defined as the comprehensive application of AI technology across all domains of warfare to achieve "algorithmic superiority" over the enemy. China has explicitly included the development of intelligent weapons and equipment in its 14th Five-Year Plan (2021-2025) and declared its intention to leverage AI as a strategic opportunity to "leap forward to become a world military power."[29] In terms of technological development, investments have been expanded in areas such as machine learning, autonomous control, and human-machine teaming through civil-military integration. Collaboration with institutions like the National University of Defense Technology (NUDT), a leader in AI research, and major civilian tech companies has been actively encouraged. Organizationally, an intelligentization promotion unit has been established under the Joint Staff Department, and pilot projects for AI utilization are being operated within each service branch, including the Army, Navy, Air Force, Rocket Force, and Strategic Support Force.[30]

For example, in 2020, the Equipment Development Department of the Central Military Commission organized an "Integrated Command Algorithm Competition." This competition involved developing reconnaissance, electronic warfare, and firepower strike plans for a simulated island landing scenario using AI, which is considered to have been conducted with the Taiwan scenario in mind.[31] Through such simulated combat exercises, the PLA appears to have been testing the tactical decision-making capabilities of AI algorithms and exploring methods of cooperation with human commanders. Significant progress has been made in terms of theory and concepts through the promotion of the intelligentization strategy. A U.S. Department of Defense report assesses that "PLA strategists believe that new technologies such as AI will provide cognitive superiority in future warfare by enabling vast information processing and machine-speed decision-making."[32]Furthermore, the Chinese military is researching concepts such as intelligent swarm attrition warfare, AI-based space warfare, and cognitive control operations as intelligent operational concepts,” and is envisioning battlefield networking through unmanned-unmanned mixed systems and distributed ISR.

[33]However, in terms of actual implementation, it is still considered to be in the early stages. Internal Chinese military documents also point to concerns such as “gaps in the military application of AI technology compared to the US military,” insufficient data quality and network security, and lack of AI reliability.[34]In other words, while intelligent strategy is conceptually established and extensive experiments are underway, experts assess that significant challenges remain in fully reflecting this in actual military power. For example, according to Chinese media, AI algorithms applied to modern warfare have derived tactical solutions through millions of simulations, and some AI decision support functions have been applied to command post exercises. However, there is little public information on whether AI command and control systems have been deployed in actual combat. Chinese military officials also express concerns about AI reliability, reportedly pointing out that “we cannot entrust our lives to unreliable AI” and that “AI malfunctions could lead to friendly casualties or unintended escalation risks.”

“Drone swarms” are an area in which China has concentrated investment to enhance its asymmetric capabilities, involving a concept where numerous unmanned aerial vehicles autonomously cooperate to encircle and attack targets or create disruptions. Through drone swarms, the Chinese military aims to overwhelm enemy air defense systems, reduce risks to manned forces, and maximize cost-effectiveness with a large number of expendable drones.[36]In terms of technological progress, China has showcased several record-breaking demonstrations in recent years. In 2018, China Electronics Technology Group Corporation (CETC) succeeded in a swarm flight of 200 fixed-wing drones, setting a world record at the time. In 2020, they released a test video showing a light tactical vehicle equipped with a 48-cell drone launcher simultaneously launching 48 small suicide drones.[37]These small loitering munitions detect targets using cameras before attacking, and are evaluated as an intermediate capability between cruise missiles and unmanned attack aircraft.

However, the operational deployment of these swarm drone technologies remains unclear. Chinese military experts acknowledge that it is uncertain whether these systems have been actually deployed by the PLA and that technical issues such as “AI response time delays” and “vulnerability to electronic jamming” still need to be resolved.[38]For instance, there was an example of operating a small drone formation during a Tibet Military Command exercise in 2020, but this was reportedly for reconnaissance and logistics purposes and does not appear to have led to the operationalization of large-scale swarm attack capabilities.[39]Nevertheless, it is clear that China is accelerating the development of swarm drones through “military-civil fusion.” The “Unmanned Competition” (无人争锋), hosted by the Air Force Engineering University, saw university and corporate teams compete in drone swarm navigation algorithms, selecting outstanding entries. Research institutes affiliated with CETC are reportedly developing smart swarm control software.[40]In summary, while China's swarm drone technology has achieved world-leading advancements from the demonstration stage, it is difficult to confirm whether it has been converted into reliably operational forces, and many opinions suggest it will take several more years. If signs of swarm drone deployment emerge in areas like the Taiwan Strait in the future, it could indicate China's successful operationalization; however, for now, it can be assessed as being in the experimental operational phase.

IV. Comparative Evaluation: Intersection of Theoretical and Practical Assessments of Innovation Speed

While both the United States and China have strategically pursued AI-driven military innovation over the past decade, there appear to be differences in their implementation patterns and force deployment achievements. In the case of the United States, AI applications for specific, limited missions, such as Project Maven, are assessed to have been absorbed into the organization and operationalized, yielding practical effects. However, innovations requiring the establishment of broad integrated systems, such as JADC2, appear to be limited to partial progress amidst a complex interplay of technology, organization, and budget.[41]Meanwhile, China is accelerating conceptualization and technological experimentation by pursuing a military-wide AI vision led by the top leadership. In some areas, particularly drone swarms, it even appears to be surpassing the United States in demonstration capabilities. However, both countries share the commonality that it is still too early to consider AI systems as core forces that will determine the outcome of war, due to a lack of combat experience and limitations in technological reliability.

Ultimately, the final test of innovation is actual combat, and considering the high degree of risk and uncertainty inherent in warfare, the “fog of war,” it appears that neither the US nor China can confidently answer the question of “has it reached a level usable on the battlefield?” While radical innovation theory predicted innovative patterns in some projects, and the US-China competition and frequent regional conflicts have indeed accelerated this, aspects of delay and stagnation, as pointed out by incremental innovation theory, are also widely observed. Regional conflicts, such as the war in Ukraine, are serving as testing grounds for innovation, but they also demonstrate the persistence and importance of traditional operational patterns.

The United States, observing the war in Ukraine, learned the importance of drones and numerous low-cost platforms, driving the Replicator initiative. However, it also realized that “logistics and production capacity in large-scale conventional warfare are decisive.” China, analyzing the experiences in Ukraine, recognizes that AI weapons did not prove decisive in high-intensity regional conflicts and appears to be emphasizing the integration of human command and conventional firepower in its envisioned intelligent warfare. This aligns with incremental innovation theory, as radical projections are modified upon encountering real-world friction. In other words, both countries are actively exploring operational innovation through AI while also adjusting their expectations based on current military realities and lessons learned. This duality is the essence of the current situation and suggests that future warfare will likely be a “complex amalgam of innovation and continuity” for a considerable period.

Despite the practical limitations of this study, one point appears relatively clear. In a situation of intense military competition, militaries are well aware of the potential effects of AI-based technologies and the significant risks that could arise if the adversary fully utilizes them. In other words, there is no “leeway” to neglect the transition to future warfare. However, it is still considered too early to significantly replace existing warfare systems with new technologies.

Therefore, efforts are proceeding in two directions. First, efforts to build highly reliable, advanced new warfare systems will continue, which will take considerable time. Second, low-level asymmetric countermeasures will be prepared to minimize damage from low-level, asymmetric attacks by the adversary, even if the countermeasures themselves are not fully developed. In essence, future warfare elements are more likely to concentrate on attempts to use them asymmetrically and counter-attempts to neutralize them, rather than immediately changing the overall battlefield. Future defense and military planning must necessarily consider this multi-faceted planning framework that connects and integrates the present and the future.■

[1]Park Sang-seop, *The History of Technology and Warfare* (Seoul: Acanet, 2018).

[2]Darrell M. West and John R. Allen, Turning Point: Policymaking in the Era of Artificial Intelligence, Washington, DC: Brookings Institution Press, 2020; Priyesh Mishra et al., Code, Command, and Conflict: Charting the Future of Military AI, Cambridge, MA: Belfer Center, Harvard University, 2025.

[3]Radha Iyengar Plumb and Michael C. Horowitz, “Is the Pentagon Slowing Artificial Intelligence Adoption?” Council on Foreign Relations, August 21, 2025.

[4]Zachary Burdette, Dwight Phillips, Jacob L. Heim, Edward Geist, David R. Frelinger, Chad Heitzenrater, and Karl P. Mueller, An AI Revolution in Military Affairs? How Artificial Intelligence Could Reshape Future Warfare, Santa Monica, CA: RAND Corporation, June 2025.

[5]Joshua Glonek, “The Coming Military AI Revolution,” Military Review, May–June 2024.

[6]Zachary Burdette et al., 2025, p. 2.

[7]Joshua Glonek, 2024, pp. 91-92.

[8]David Kirichenko, “How AI Is Eroding the Norms of War: An unchecked autonomous arms race is eroding rules that distinguish civilians from combatants.” AI Frontiers, May 27, 2025.

[9]Barry R. Posen, The Sources of Military Doctrine: France, Britain, and Germany Between the World Wars, Ithaca, NY: Cornell University Press, 1984; Stephen Peter Rosen, Winning the Next War: Innovation and the Modern Military, Ithaca, NY: Cornell University Press, 1991.

[10]Michael C. Horowitz, The Diffusion of Military Power: Causes and Consequences for International Politics, Princeton, NJ: Princeton University Press, 2010.

[11]Lance Menthe, Li Ang Zhang, Edward Geist, Joshua Steier, Aaron B. Frank, Erik Van Hegewald, Gary J. Briggs, Keller Scholl, Yusuf Ashpari, Anthony Jacques, Understanding the Limits of Artificial Intelligence for Warfighters, Volume 1, Summary, Santa Monica, CA: RAND Corporation, January 2024.

[12]UK Parliament Parliamentary Office of Science and Technology, Electromagnetic (Electronic) Warfare, London: POST, July 10, 2025.

[13]Mary E. Oakley, Weapon Systems: Key Aspects of DOD’s Capability Development and Acquisition Process, testimony before the Subcommittee on Seapower and Projection Forces, Committee on Armed Services, House of Representatives, GAO-25-107928, Washington, DC: U.S. Government Accountability Office, May 21, 2025.

[14]Dominika Kunertova, Tomorrow’s Drone Warfare, Today’s Innovation Challenge: Learning from the Ukrainian Battlefield, Zurich: Center for Security Studies, ETH Zurich, May 2024.

[15]Given the general-purpose nature and innovativeness of artificial intelligence technology, it is not easy to predict how this technology will change the way war is waged and the balance of military power among nations. However, major military powers, including the United States, appear to have concretized their policy directions for military innovation and defense reform in preparation for the anticipated future war competition. Artificial intelligence can also revolutionize the way nuclear weapons are operated, including nuclear command and control, and this has the potential to alter the nuclear balance and military balance among major powers. However, this study limits its analysis to conventional weapon systems and battlefields, excluding nuclear weapons, from the perspective that the balance of strategic weapons will be maintained through countermeasures by all nations, considering their ripple effects, and thus an overall balance will be achieved. Reference: Seol In-hyo, “Future Warfare and the Construction of Data-Based Intelligent Integrated Systems: Focusing on the Case Analysis of U.S. Joint All-Domain Command and Control (JADC2),” 『Journal of Strategic Studies』 31 (3), 2024.

[16]Seol In-hyo and Park Won-gon, “Prospects for U.S. Defense Strategy under the New Administration and Implications for the ROK-U.S. Alliance: Focusing on the Acceptance and Transformation Potential of the ‘Third Offset Strategy’,” 『Korea Defense Policy Review』 115 (1), 2017.

[17]Yoon Daeyeop, “Defense Artificial Intelligence Transformation and Digital Military Innovation: Lessons from Project Maven and the Russo-Ukrainian War,” 『Korean Political Science Review』 34 (2), 2025.

[18]Jaspreet Gill, “NGA making ‘significant advances’ months into AI-focused Project Maven takeover,” Breaking Defense, May 24, 2023.

[19]Mina Al-Oraibi, “US used AI to find targets for strikes on Syria and Yemen,” The National (relying on Bloomberg report), February 26, 2024.

[20]Billy Mitchell, “NATO inks deal with Palantir for Maven AI system,” DefenseScoop, April 14, 2025.

[21]Jon Harper, “US military deploys new JADC2 capability to Middle East,” DefenseScoop, April 3, 2024.

[22]U.S. Government Accountability Office (GAO), Defense Command and Control: Further Progress Hinges on Establishing a Comprehensive Framework, GAO-25-106454, April 8, 2025, pp. 1–3.

[23]Congressional Research Service, DOD Replicator Initiative: Background and Issues for Congress, IF12611, updated November 2025, p. 2.

[24]Patrick Tucker, “First Replicator drones already in Indo-Pacific, DOD says,” Defense One, May 23, 2024.

[25]Congressional Research Service, 2025.

[26]Congressional Research Service, 2025.

[27]Brandi Vincent, “In wake of Project Maven, Pentagon urged to launch new ‘pathfinder’ initiatives to accelerate AI,” DefenseScoop, July 18, 2023.

[28]Office of the Secretary of Defense, Military and Security Developments Involving the PRC 2023, Department of Defense report, October 2023, pp. 114–115.

[29]Sam Bresnick, China’s Military AI Roadblocks: PRC Perspectives on Technological Challenges to Intelligentized Warfare, CSET Reports, June 2024.

[30]Marcus Clay, “The PLA’s AI Competitions: Can the new design contests foster a culture of military innovation in China?” The Diplomat, November 5, 2020.

[31]Marcus Clay, 2020.

[32]Office of the Secretary of Defense, 2023, p. 97.

[33]Office of the Secretary of Defense, 2023, p. 97.

[34]Sam Bresnick, 2024, pp. 1-2.

[35]Sam Bresnick, 2024, pp. 29-30.

[36]John Harper, “US military deploys new JADC2 capability to Middle East,” DefenseScoop, April 3, 2024.

[37]Ryan Pickrell, “China is practicing unleashing swarms of suicide drones packed with explosives from the backs of trucks,” Business Insider, October 16, 2020.

[38]Ryan Pickrell, 2020.

[39]Antoine Bondaz & Simon Berthault, “China’s use of drones in the Sino-Indian border dispute: a concrete example of civil-military integration,” Foundation for Strategic Research Report, July 2023.

[40]Marcus Clay, 2020.

[41]U.S. Government Accountability Office (GAO), 2025, p. 11.

■Author: Seol In-hyo_Professor, Department of Strategic Studies, Korea National Defense University.

■ Editor: Lim Jae-hyun_EAI Research Fellow

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