[Choosing a Path in the AI Era and a New Korean Peninsula Initiative] ① Changes in the Strategic Environment of the AI Era and a Future Korean Peninsula Initiative

I. Background and Purpose of the Study[1]

For the past three decades, South Korea’s policy toward North Korea has been based on an understanding of denuclearization, inter-Korean exchange and cooperation, the establishment of a peace regime, and long-term unification as interconnected processes. This approach was rooted in the expectation that the North Korean nuclear issue could be resolved through negotiation and compensation, that expanded inter-Korean exchange would foster change in North Korea, and that improved U.S.-North Korea and inter-Korean relations would together pave the way for peace and unification on the peninsula. While conservative and progressive administrations have differed on the methods for denuclearization, the pace of inter-Korean cooperation, and the relationship between a peace regime and the alliance, they have shared significant common ground in viewing denuclearization, peace, cooperation, and unification as linked, long-term objectives.

The Korean Peninsula of the mid-2020s is departing significantly from the conditions that shaped this earlier vision. By enacting its law on nuclear force policy, institutionalizing the operation of tactical nuclear weapons, and advancing its nuclear delivery capabilities, including ICBMs and SLBMs, North Korea is cementing nuclear weapons as a permanent component of its national strategy and regime security, rather than as a temporary bargaining chip. Pyongyang’s perception of inter-Korean relations has also shifted. Its redefinition of the relationship as one between “two hostile states”—a departure from its long-held position of viewing South Korea as a partner for unification within a single nation—cannot be explained as a mere adjustment in its policy toward the South. This shift reflects a simultaneous transformation in its national identity, its perception of unification, its view of the South, and its foreign strategy. Moreover, the international conditions that supported the previous approach of achieving denuclearization and improved relations through phased negotiation and engagement are eroding. This is due to a confluence of factors: the strengthening of North Korea-Russia strategic ties, the structural nature of U.S.-China competition, China’s preference for the status quo on the peninsula, and a similar status-quo orientation in U.S. policy toward North Korea.

This change does not mean that all premises of existing North Korea policy must be discarded. Denuclearization, peace, exchange, cooperation, and unification remain important goals for South Korea to pursue, and peaceful coexistence still offers a realistic direction for managing the current state of inter-Korean relations. The problem is that the causal relationships that once connected these goals can no longer be assumed to function as they did before. If North Korea’s nuclear capabilities persist for an extended period and the “two hostile states” relationship becomes institutionalized while inter-Korean exchange and political dialogue remain severed, a stable peace will not emerge simply because war does not break out. If current trends continue long-term, North Korea’s nuclear status could become a de facto constant, inter-Korean hostility could become institutionalized, and unification could be relegated to a constitutional and normative goal increasingly detached from practical policy. The strategy toward North Korea for the next 20 to 30 years must not only reiterate existing goals but also address how to institutionalize peaceful coexistence and maintain the possibility of long-term integration under these changed conditions.

Artificial intelligence is a new variable converging with this structural transformation on the Korean Peninsula. It is difficult to view AI as merely an advanced industry or a specific technological field because it affects the entire process of how humans and organizations handle information—from collection, classification, analysis, and prediction to content generation, decision support, automation, simulation, and communication. Large language models, multimodal models, generative AI, and autonomous agents are combining with robotics, sensor networks, cloud infrastructure, and advanced semiconductors to transform not only corporate production methods but also national administration, knowledge production, military operational methods, and society’s information environment. The strategic importance of AI lies less in the performance of individual technologies and more in its ability to change how nations and societies produce and interpret information and weigh future options.

The impact of AI on international politics also stems from this characteristic. The competition between the United States and China is expanding to encompass not only semiconductors and advanced computing but also data, the cloud, talent, algorithms, standards, and military applications. Nations that effectively leverage AI have the potential to secure new advantages in production, finance, supply chains, R&D, education, healthcare, and the energy transition, while countries with limited access to technology and infrastructure may face development gaps and technological dependency. The ways in which states manage citizens and information and provide administrative services are also being influenced by AI. Furthermore, as AI is integrated into surveillance and reconnaissance, targeting, unmanned systems, cyber operations, electronic warfare, and command and control, the composition of military forces and the methods of warfare are changing.

Many recent changes in the international order are not proceeding in isolation from AI development. The weakening of the liberal international order, U.S.-China strategic competition, the bloc-formation of technology supply chains, the securitization of semiconductors and data, the military AI race, the spread of digital authoritarianism, and the technological divide with the Global South are all intertwined. AI is both affected by these changes and acts as a variable that accelerates them. The technological and production networks once built for efficiency under open globalization are gradually being reconfigured according to the logic of security, sovereignty, and trusted supply chains. In an international environment where technological bloc-formation, strategic decoupling, controlled interdependence, data sovereignty, and standards competition are all underway, AI is not just one arena of great power competition but a primary domain in which the international order is being reshaped.

The Korean Peninsula is already a space where military confrontation, the nuclear issue, systemic competition, alliance politics, and great power rivalry overlap. The development of AI adds a new dimension of competition and dependence to this mix. Due to constraints in high-performance semiconductors, large-scale computing infrastructure, data, and a commercial tech ecosystem, it will be difficult for North Korea to independently develop frontier AI. However, this limitation does not eliminate North Korea from AI development. It can selectively utilize AI for specific purposes by using open-source models, its own personnel, and its cyber capabilities, and it is likely to become partially dependent on China's open-source models and technological ecosystem. The strengthening of its strategic partnership with Russia could also provide new technological pathways for North Korea's asymmetric military capabilities, such as drones, electronic warfare, and cognitive warfare. Since South Korea is also developing its AI-based capabilities in surveillance and reconnaissance, intelligence analysis, precision strikes, and crisis management, the military application of AI by both Koreas has the potential to make the existing security dilemma faster and more complex.

A more fundamental long-term issue than military competition may be which AI ecosystem North Korea becomes integrated into. If North Korea, facing limits to its independent technological development, continuously relies on China's platforms, hardware, data processing methods, technical standards, and digital governance models, its AI development could evolve beyond simple technology adoption into a structural dependence on a specific technological order. If North Korea's data sovereignty and technological development trajectory become locked into a Chinese-style ecosystem, and this is combined with surveillance technology and social management models, the digital authoritarian nature of the North Korean regime could be reinforced. If its military and cyber AI become closely linked to China's tech ecosystem, the security of the Korean Peninsula will become difficult to address solely as an inter-Korean issue and could be subsumed into the U.S.-China technological and military competition.

The impact of these changes on long-term inter-Korean relations needs to be viewed more broadly than just the widening technology gap. In the AI era, knowledge production and economic development are heavily influenced by connectivity to data, research networks, international standards, and technological ecosystems. If North Korea's technological isolation and dependence on a specific country accumulate over a long period of severed inter-Korean cooperation, not only economic productivity but also differences in knowledge systems and governance could widen. Social resources that have maintained the traditional national community, such as separated families, language, and culture, could also weaken over time. If the present is evaluated solely based on the possibility of political unification, these long-term changes may not be fully apparent. The more the technological, knowledge-based, institutional, and social connections that make integration actually possible are weakened, the greater the costs of future integration could become.

AI also offers possibilities in the opposite direction. In fields with relatively low risk of military diversion that directly contribute to the human security of the North Korean people—such as infectious diseases, disasters, climate, food, forestry, water resources, and public health—it is possible to consider forms of technological cooperation different from past inter-Korean efforts. The organization and connection of data on separated families, the accumulation of inter-Korean language and educational materials, and the preservation of cultural heritage and analysis of humanitarian information are also areas that could be considered in the long term. Based on its experience in digital development cooperation and sovereign AI, South Korea can explore cooperation models suited to North Korea's language, geography, climate, and development conditions, which could offer an alternative to North Korea's reliance on a single external technological ecosystem.

The potential for cooperation should not be overstated. AI is a technology with strong dual-use characteristics, and even widely used civilian technologies like data analysis or image recognition can be repurposed for military, cyber, and surveillance objectives. AI cooperation with North Korea must not strengthen its military or social surveillance capabilities. It is necessary to design the objectives of cooperation, the scope of data, the users of the technology, verification methods, and post-monitoring systems together. Rather than an approach that expands technology transfer simply because inter-Korean relations improve, a more realistic path is risk management-based cooperation that distinguishes between areas where cooperation is possible and areas where it must be restricted, proceeding step-by-step from fields that contribute to human security and have a low risk of diversion.

The new unification initiative proposed by this study takes this duality as its starting point. If unification is understood only as a single moment of political agreement or institutional integration, it is difficult to fully grasp the long-term impact of AI on inter-Korean relations. Regardless of how the political relationship unfolds, the actual process of connecting administration, economy, health, education, knowledge, data, and social trust will take a long time. If unification is viewed as a long-term process of integration that includes technology, data, knowledge, governance, and human security, it becomes possible to analyze how current cooperation and risk management shape the conditions for future integration.

The framework for North Korea policy is also difficult to set as a simple choice between pressure and engagement. This is because AI must be utilized for deterrence and defense, while it can also be used for crisis management and human security cooperation. A policy to deter and manage the risks of North Korea's military and cyber AI development is not mutually exclusive with a policy that prepares for the possibility of cooperation in areas where conditions permit. A long-term unification initiative also needs to be situated within the context of U.S.-China AI competition, the bloc-formation of the international order, the technological divide with the Global South, and the formation of international AI norms and technical standards, rather than being explained solely through the internal inter-Korean relationship.

It is difficult to conduct this analysis based on technological determinism that assumes AI will promote unification. If North Korea utilizes AI for its military and social control, deepens its dependence on China's tech ecosystem, and the knowledge and technology gap between the two Koreas widens, AI could instead more stably reproduce division and increase the costs of future integration. Because of this possibility, it is also difficult to leave AI solely as an industrial and technological policy separate from unification policy. South Korea needs to proactively analyze the direction of change, manage the military and social risks, and simultaneously prepare for the possibility of limited cooperation that can be utilized when political conditions allow. A new unification initiative does not presuppose uncertain AI development as an automatic driver of unification; rather, it is an initiative that seeks to explore, among the various paths AI might create, the conditions that can expand peace and the possibility of long-term integration on the Korean Peninsula.

II. Key Research Questions and Scope

The first reason this study includes AI in its vision for Korean unification lies in the changing military security environment. The existing security dilemma on the Korean Peninsula has been shaped by the interplay of North Korea's nuclear and missile capabilities, the deterrence posture of South Korea and the United States, conventional military forces and exercises, and alliance politics. When AI is integrated into military systems, the basic structure of this dilemma may remain, but the speed of information processing and operations, the perception of the adversary's capabilities, and the relationship between offense and defense can change. While AI-based surveillance and reconnaissance can contribute to more accurately detecting an adversary's military activities, the adversary may perceive this as an enhancement of preemptive strike capabilities. Advances in missile defense and target identification capabilities, while strengthening deterrence for the defending side, may be perceived by North Korea as a change that weakens the survivability of its nuclear forces. Automated cyber responses could increase the risk of escalation if an error occurs in judging the attacker. On the Korean Peninsula, with its short geographical depth and the involvement of nuclear weapons, this increase in decision speed and crisis instability can have greater strategic significance than in other regions.

The second issue is the development path of North Korean AI. North Korea lacks high-performance semiconductors, large-scale computing infrastructure, sufficient data, global research networks, a commercial AI ecosystem, and open academic exchange. Sanctions also constrain its technological development. It seems unlikely that North Korea can independently build general-purpose AI capabilities comparable to those of frontier companies in the United States, China, or Europe. However, the limits of indigenous development do not eliminate the possibility of using AI in a limited capacity. North Korea can selectively utilize purpose-built AI in military domains like drones, image analysis, cyber operations, and surveillance; in economic areas such as production and logistics management; and in knowledge and governance sectors including education, propaganda, and social surveillance. If its dependence on China's AI ecosystem expands in this process, North Korea's technological development will face not only the challenge of supplementing its capabilities but also the problems of data sovereignty, technological standards, and dependency on a digital governance model. This is why North Korea's AI problem must be analyzed not only by its own technological level but also in terms of which external technological ecosystem it integrates with.

The third research question stems from the idea that the form of inter-Korean cooperation may change in the AI era. In the past, inter-Korean cooperation centered on large-scale economic projects, infrastructure connectivity, humanitarian aid, socio-cultural exchanges, and military tension reduction. Under the current circumstances—where North Korea's “two hostile states” doctrine, the severing of inter-Korean communication channels, the sanctions regime, the uncertainty of U.S.-North Korea relations, and the bloc-formation of the international order persist—it is difficult to restore past cooperation methods as they were. Cooperation using AI has the potential to start on a relatively small scale, with clear technical objectives and verifiable results, in areas directly related to the human security of the North Korean people. Fields such as disaster response, public health, food, and forestry fall into this category. However, considering the dual-use nature of AI and the potential for its diversion to surveillance and governance, cooperation itself cannot always be seen as having a positive effect. Evaluating the necessity and risks of cooperation simultaneously and pre-defining the scope of technology transfer, verification, and end-use becomes a crucial condition for inter-Korean cooperation in the AI era.

The fourth issue is preparation for future integration. Even if an agreement on political unification is reached, the integration of administration, law, finance, health, education, industry, infrastructure, welfare, land and resident registration, regional development, and socio-psychological aspects will not happen simultaneously. The longer the division and development gap persist, the larger the scale of information, institutions, professional personnel, and social costs required for actual integration can become. AI can be used to analyze in advance, based on data, the vulnerabilities in North Korea's infrastructure, health, and nutrition; its educational gaps and industrial transition needs; environmental restoration and urban planning; energy grids and population movements; and fiscal demands, and to review various integration scenarios. Even in such applications, efficiency cannot be the sole criterion. The use of data must consider the protection of personal information and human dignity, and the efficiency of administrative integration must be addressed alongside the formation of social trust.

The scope of this study is set according to these four research questions. First, it reviews global trends in AI development and the changes in the international order surrounding AI. Then, it analyzes the conditions under which North Korea is developing and utilizing AI, dividing the analysis into the military, economic, and knowledge/governance domains. In each domain, it not only assesses the risks of North Korean AI but also examines the limited opportunities it may present for inter-Korean relations. North Korea's future AI trajectory is examined by comparing the possibilities of indigenous development, cooperation with China, and inter-Korean cooperation. Finally, it presents the direction for South Korea's AI cooperation with the North and risk management strategies, as well as the implications for a long-term unification initiative. This study does not aim to evaluate the detailed engineering characteristics of AI or the performance of individual models; rather, it analyzes at a strategic level how technological change can alter the conditions for security, economy, knowledge, governance, and long-term integration on the Korean Peninsula.

III. Structure and Analytical Framework of the Study

The study is structured to move from the international environment shaped by AI to an analysis of North Korea's choices and their impact on inter-Korean relations, and finally to South Korea's response.

Chapter 2 deals with the global development of AI and changes in the international order. It examines how AI is proliferating in the military, economic, and knowledge/governance domains, while also reviewing the U.S.-China AI competition, semiconductor supply chains, data sovereignty, digital authoritarianism, and the AI governance debate. The reason for first analyzing the international environment of technological bloc-formation and strategic decoupling is that North Korea's AI development cannot be explained solely by its internal technology policies.

Chapter 3 analyzes North Korea's AI policies, technological development, and educational and research infrastructure, and examines how AI can be utilized in the military, economic, and knowledge/governance sectors. It looks at the extent to which North Korea can develop its own AI capabilities with limited resources, as well as the possibility of its growing dependence on China's technological ecosystem. The subsequent three chapters analyze this issue by domain.

Chapter 4 addresses military AI, crisis instability on the Korean Peninsula, and the issue of strategic stability, and examines the possibility of managed military competition in response to North Korea's AI militarization.

Chapter 5 analyzes the areas where AI can be utilized in the North Korean economy and the opportunities and constraints for inter-Korean economic cooperation, comparing the different paths of cooperation with China and inter-Korean cooperation.

Chapter 6 examines the impact of AI on North Korea's governance, surveillance, social management, education, and knowledge production, and reviews the possibilities and limitations of a Korean Peninsula AI Commons that could accumulate shareable knowledge and data resources in the long term.

Chapter 7 synthesizes the preceding sectoral discussions to present the direction for South Korea's AI cooperation with North Korea and the implications for a long-term unification initiative.

The concluding Chapter 8 sets out different future paths for North Korea—subordination to China, indigenous development, and seeking cooperation with South Korea—and compares how the military, economic, and knowledge/political domains might unfold along each path. The purpose of this scenario-based analysis is not to predict any single future, but to distinguish which risks South Korea's policies should prioritize managing and which cooperation possibilities it should prepare for under different conditions.

The first analytical axis running through this structure is the relationship between technology, security, and peace. The political effects of AI are not fixed within the technology itself. The same image analysis and information processing technology can heighten an adversary's threat perception if combined with surveillance, reconnaissance, and preemptive strikes, but it also has the potential to contribute to military stability if used for early warning, crisis management, and armistice monitoring. To evaluate the effect of technological development on peace on the Korean Peninsula, one must look not only at the purpose for which AI is used but also at the military and political context in which the technology is situated.

The second analytical axis is the simultaneity of risk and opportunity. North Korea's AI development carries the risk of strengthening its military power, cyber capabilities, population surveillance, and social control, but it also offers the potential for new means of cooperation in areas such as disaster response, public health, food, and human security. Similarly, the AI-based military capabilities developed by South Korea can have the dual effect of enhancing deterrence against North Korea and deepening crisis instability. Rather than pre-defining AI as either a technology that contributes to peace or one that deepens division, it is necessary to compare the different effects that may arise depending on the sector, purpose of use, and international and institutional conditions.

The third analytical axis is a comparison of the paths North Korea can choose. Indigenous development is the path that allows North Korea the greatest autonomy, but it faces significant constraints in terms of necessary computing infrastructure, data, and research ecosystems. Cooperation with China is a realistic way to partially offset these constraints, but it also entails the possibility of long-term technological subordination and the deepening of digital authoritarianism. Inter-Korean cooperation has the potential to contribute to human security, economic development, and the formation of long-term integration capacity, but it is difficult to realize without resolving issues of military diversion, surveillance and control, sanctions, and technology security. Through a comparison of these three paths, this study aims not to assert which choice North Korea will make, but to analyze the costs and possibilities that each path brings for the security of the Korean Peninsula, North Korea's development, and long-term inter-Korean relations.

The normative criteria this study uses to evaluate these paths are set accordingly. The use of AI should, first and foremost, not undermine military stability on the Korean Peninsula. Possible cooperation should contribute to improving the human security of the North Korean people. The effect of maintaining or expanding trust and long-term integration capacity between the two Koreas also needs to be considered. Technological cooperation that increases military diversion, population surveillance, human rights violations, and cyber threats cannot be seen as contributing to long-term peace and integration, so the institutional conditions for restricting and verifying such cooperation also become part of the evaluation.

A new unification initiative for the AI era does not treat unification solely as a single political event that will occur in the distant future. The more uncertain the timing of political unification, the more important it may become in the long run what capabilities, institutions, knowledge, and connections are maintained and accumulated. In a situation where inter-Korean exchange has virtually ceased, the “two hostile states” relationship is strengthening, and the international order is forming into blocs, it is difficult to expect short-term results from cooperation. South Korea's policy must continue deterrence and defense while managing the military and social risks that AI will bring, and prepare for the possibility of human security cooperation and knowledge-based and institutional connections that can be utilized when the political environment changes.

AI is not a technology that will automatically solve the existing problems of the Korean Peninsula, nor is it simply another variable added to the already established structure of division. Depending on how the direction of technological development, North Korea's choices, the evolution of U.S.-China competition, and the political conditions of inter-Korean relations combine, AI could more stably reproduce hostility and subordination, or it could be used as a means to prepare for limited cooperation and long-term integration. The subsequent analysis will examine in detail the conditions under which these different possibilities become reality. ■

[1]This working paper was written with the support of the “2025 Research Support Program of the Unification and Nanum Foundation (in the fields of unification studies, North Korean studies, and inter-Korean integration policy).”

■ Author: Jaesung Chon_President of EAI, Professor, Department of Political Science and International Relations, Seoul National University.

■ Managing Editor: Jaehyun Lim_Research Fellow, EAI
For inquiries: 02-2277-0746 (ext. 209) jhim@eai.or.kr