[International Politics in the Age of AI] ⑨ AI and International Political Economy: National AI Strategies and Global Competition

⑩ AI and International Political Economy, Jiyeon Song [Read Working Paper]

I. Introduction

In the age of artificial intelligence (AI), AI is emerging as a key strategic asset in national competition and international political economy. The rapid development and diffusion of AI technology are evaluated as technologies with the potential to reshape core elements of national power, including economic growth, military strength, data control, and financial markets, beyond mere industrial innovation. Against this backdrop, countries are competitively formulating and announcing National AI Strategies, leading to intense competition to secure global technological leadership.

This study first systematically analyzes these National AI Strategies by classifying them into four ideal types. First, based on the goal orientation of the strategy (offensive vs. defensive), it distinguishes whether AI will be used as a means to secure global hegemony or to manage the risks and dependencies arising from technological diffusion. Second, based on the policy leadership structure (market-led vs. state-led), it determines whether innovation will be left to private companies and the market, or whether the state will directly coordinate industry, data, and infrastructure. Combining these two criteria yields four strategic types: 'market-led-offensive,' 'state-led-offensive,' 'market-coordinated-defensive,' and 'state-led-defensive.' Each country adopts different strategic choices according to its political system, industrial structure, technological ecosystem, and international standing.

Furthermore, National AI Strategies must be understood as dynamic products interacting within the structural context of global AI competition. As a general-purpose technology, AI has ripple effects across economic, industrial, military, and social domains, and the technological gap between nations directly translates into a reshaping of national power and strategic standing, beyond mere productivity differences. Therefore, AI tends to intensify global competition into an all-out confrontation. However, this AI competition carries the risk of amplifying the negative effects of the strategies pursued by each country. Thus, establishing international AI governance is an urgent task to manage these risks.

II. National AI Strategies

As the strategic importance of AI has rapidly increased recently, countries have been competitively formulating and announcing 'National AI Strategies.' In March 2017, the Canadian government announced the 'Pan-Canadian Artificial Intelligence Strategy,' accompanied by comprehensive financial support. This initiative defined AI not merely as an industrial technology but as a strategic technology that would determine national competitiveness, economic growth, and long-term innovation capabilities, and represented an attempt to systematically coordinate investment in research and development, talent cultivation, and the building of an innovation ecosystem at the national level.[1]Subsequently, as numerous countries announced their AI strategies, national-level AI policies have spread globally. The emergence and proliferation of these National AI Strategies demonstrate that AI is recognized not just as a new technology but as a general-purpose technology capable of reshaping core elements of national power, including productivity, industrial structure, military strength, data control, and financial markets. Therefore, National AI Strategies need to be understood not as a simple expansion of technology policy but as a product of strategic responses by nations to redefine their political and economic positions amidst technological hegemony competition and the reshaping of global value chains.[2]

National AI Strategies can be defined as 'a set of coordinated government policies under clear objectives to maximize the potential benefits and minimize the potential costs that AI brings to the economy and society.'[3]This definition encompasses three core elements. First, AI strategy presupposes the setting of clear objectives at the national level. Second, various policy instruments, including R&D, industrial policy, education and workforce development, regulation, data governance, and international cooperation, are combined in a coordinated manner. Third, both offensive goals of promoting growth and defensive goals of managing risks are considered. In practice, governments have set different policy objectives, such as securing global AI leadership, strengthening technological sovereignty, protecting strategic industries, and overseeing and regulating AI adoption. This implies that National AI Strategies are not merely industrial promotion policies but comprehensive blueprints for how the state will manage and direct the development and diffusion of AI technology within a given political and economic order.

This paper presents four 'ideal types' of AI development strategies to systematically understand these National AI Strategies. The ideal type, a concept proposed by Max Weber, is not an empirical reality that exists exactly as is but a conceptual model constructed to analytically grasp complex and mixed realities.[4]This is not intended for simple classification of reality but provides analytical benchmarks for comparing and evaluating how each national strategy more closely aligns with certain directions and institutional characteristics. While actual National AI Strategies vary depending on each country's political system, industrial structure, technological ecosystem, financial system, and social value system, they can be typologized based on their fundamental direction and policy leadership structure.

The first criterion for distinguishing ideal types is 'strategic goal orientation.' This refers to the ultimate values prioritized by the AI strategy and the role AI plays as a means within the national power and development model. Here, it can be divided into offensive and defensive strategies. An offensive strategy is an approach that seeks to preemptively secure a competitive advantage in AI amidst global technological hegemony competition, thereby achieving economic, military, and scientific-technological superiority. Key tasks include accelerating technological innovation, accumulating data and computing resources, forming network effects, securing leadership in global technology standards and infrastructure, and controlling core components and supply chains. In this case, AI is considered a growth engine and a strategic asset, directly linked to the expansion of national power. In contrast, a defensive strategy focuses on minimizing risks that AI competition may bring, such as industrial dependency, data vulnerability, labor market shocks, and deepening social inequality. This does not mean suppressing technological development but rather represents a coordinated approach that seeks to secure economic autonomy and social stability by institutionally managing the speed, scope, and methods of adoption. Therefore, 'strategic goal orientation' is a conceptual axis that gauges the fundamental direction of policy: whether it focuses on actively utilizing AI as a means to expand national power or on managing the structural shocks brought about by AI diffusion.

The second criterion is 'policy leadership structure.' This refers to how authority and leadership are distributed between the state and the market, particularly digital platform companies, in the process of formulating and implementing AI strategies. Here, 'policy leadership structure' does not simply mean the level or intensity of government intervention but is a structural concept that explains which actor ultimately holds the core decision-making authority and strategic control within the governance system surrounding AI development. In market/platform-led strategies, private companies, especially large technology platform companies and venture ecosystems, act as the core drivers of innovation. The government provides the institutional framework, such as supporting basic research, refining regulatory frameworks, promoting competition, and establishing ethical guidelines, but the specific direction and speed of technological innovation are determined by market competition, capital markets, and network effects. In contrast, in state-led strategies, the government sets the priorities for AI development and directly coordinates technological advancement using industrial policy, financial support, public research institutions, public procurement, and regulatory authority. The designation of strategic industries, large-scale public investment, national control of data infrastructure, and linkage with security are characteristic features. In this case, the state functions not merely as a facilitator but as a director and a central actor in resource allocation.

Combining these two criteria yields four ideal types of National AI Strategies: 'market-led-offensive,' 'state-led-offensive,' 'market-coordinated-defensive,' and 'state-led-defensive.' Of course, actual National AI Strategies are not entirely confined to one of these types but tend to have complex characteristics and lie on a continuum. Nevertheless, these ideal types provide a useful analytical framework for analyzing how various countries are responding to the development and diffusion of the general-purpose technology known as AI.

<Table 1> Ideal Types of National AI Strategies

1. Market-Led Offensive Strategy

The market-led offensive strategy refers to a type of strategy that recognizes AI as a core strategic asset in the competition for national power, but places the primary driving force for its development not on direct state control or centralized planning, but on market competition and private innovation capabilities. In this strategy, the state functions not as a controller dictating technological development in detail, but as a coordinator and facilitator that presents a long-term vision and normative direction and designs the institutional and financial framework. The agents of innovation are decentralized actors such as corporations, universities, research institutes, venture ecosystems, and capital markets, and the state strategically enhances their competitiveness through measures such as expanding R&D investment, opening up data and computing resources, streamlining regulations, establishing standards strategies, stabilizing supply chains, and implementing selective security measures. Therefore, this strategy has the characteristics of 'strategic liberalism,' which pursues offensive goals of securing technological hegemony, preempting international standards, and achieving military superiority while advocating for a free market order.

The field of AI is an area particularly suited for this kind of strategic liberalism. Since the so-called 'AI winter' in the 1970s, when public support declined, private companies have led its development. Key innovations such as industrial robots, data mining, cloud-based AI, and large language models have primarily occurred centered around private companies. Even today, major technology companies like Microsoft, Google, Amazon, Meta, and OpenAI are leading AI development based on vast data, computing resources, and global platforms.[5] Consequently, AI is evaluated as a typical market-centric technology where inter-company competition determines the speed and direction of innovation. This private-led structure is also a key reason why the market-led offensive strategy has high consistency in the field of AI.

The United States is the most representative example of this strategy. The "American AI Initiative" announced by the Trump administration in 2019 set the maintenance and strengthening of US AI leadership as a national goal, but its approach focused on creating an innovation environment rather than centralized industrial control. Key elements included expanding federal R&D investment, opening up high-quality federal data and computing resources, minimizing regulatory barriers hindering innovation, fostering STEM talent, and participating in the formation of international norms. In particular, it sought to institutionally guarantee the autonomy of private innovation by avoiding excessive ex-ante regulation and emphasizing performance-oriented regulation based on cost-benefit analysis. This can be evaluated as a facilitative strategy that sets offensive goals while focusing on ecosystem creation and institutional foundation building in terms of policy instruments.[6]

The "Winning the Race: America’s AI Action Plan" announced by the second Trump administration in 2025 has evolved into a more explicit and structured offensive strategy. This plan is structured around three pillars: accelerating AI innovation, building US AI infrastructure, and leading international AI diplomacy and security, and it places the US-China strategic competition front and center. It includes an integrated execution strategy encompassing industry, security, and diplomacy, such as regulatory review and deregulation enhancement, promotion of open-source ecosystems, expansion of semiconductor, data center, and energy infrastructure, construction of high-security AI infrastructure for military and intelligence agencies, strengthening export controls, and forming alliance-based technology blocs. In other words, while the 2019 strategy focused on laying the institutional foundation for the innovation ecosystem, the 2025 strategy can be seen as a comprehensive execution framework that seeks to convert the fruits of innovation into geopolitical advantage through supply chain reorganization, technology control, and expansion of alliance networks. The 2025 America's AI Action Plan has further strengthened the state's strategic role through the explicit articulation of AI hegemony competition, the strengthening of AI's securitization, and the institutionalization of technology control and bloc formation strategies against competitor nations.[7] Nevertheless, both strategies can be evaluated as consistently maintaining their market-led nature, as the government does not operate the AI industry in a planned economy manner and places the center of innovation with private companies and research communities.

<Table 2> Comparison of American AI Initiative vs. America’s AI Action Plan

The strength of the market-led offensive strategy lies, above all, in its structural consistency with the private-centric development structure of AI. A competitive market structure enables experimentation with diverse technological paths, efficiently filters out failures, and promotes large-scale inflows of capital and talent. Companies with global platforms can spread international standards and technology stacks through network effects, and the state can flexibly respond to geopolitical risks by selectively intervening in strategic areas such as supply chains, security, and standards. In other words, a structural division of labor is established where the speed and scale of technological development are secured by the market, and the strategic transformation of its outcomes is handled by the state. However, this strategy also entails structural limitations. Since companies make decisions based on risk and return, long-term security, ethics, and social responsibility may be relegated to a lower priority. The strengthening of market dominance by platform companies and data monopolies can hinder competition and exacerbate inequality. Furthermore, issues of strategic consistency can arise from the fact that the commercial interests of private companies do not always align with the long-term strategies of the state.[8]

The market-led offensive strategy can be described as a hybrid strategy that seeks to secure superiority in international technological hegemony competition by mobilizing private innovation capabilities to the maximum extent, through designing institutional environments and strategic incentives, rather than through direct state control of industry. The US case demonstrates a trend toward actively pursuing the reorganization of the geopolitical order centered on technology, while maintaining an open innovation ecosystem.

2. State-Led Offensive Strategy

The state-led offensive strategy defines AI not merely as a means of industrial innovation, but as a core strategic asset directly linked to the structural reorganization of national power. It aims to secure a leading position in international technology competition by having the central government comprehensively lead long-term vision setting, resource allocation, institutional design, industrial organization, and security integration. In such a strategy, the central government leads the development roadmap, priority technologies, industrial deployment, data management systems, and the direction of standards and regulations, while markets and companies are mobilized and coordinated within the framework of national strategy.

China is the most typical example of this ideal type. Through the "A New Generation Artificial Intelligence Development Plan" released by the State Council in 2017, China elevated AI to the highest priority of national strategy and set a long-term goal of becoming a major hub and leading country in global AI innovation by 2030. This plan clearly exhibits an offensive orientation by defining AI as a strategic technology leading the future, emphasizing the domestic acquisition of strategic assets such as core algorithms, advanced AI chips, data resources, and intelligent systems, and simultaneously pursuing technological self-reliance and global competitive advantage. In particular, by directly linking AI development with military modernization through integration with the military-civil fusion strategy, it has evolved into a national project that integrates technology, industry, and security.

In terms of policy-led structure, the central government oversees the direction of development through measures such as reorganizing national key laboratories and innovation bases, injecting large-scale fiscal and policy financing, designating strategic industries and providing subsidies, and establishing data governance systems. Major platform companies such as Alibaba, Tencent, Baidu, and Huawei function as partners in the execution of national strategy, while also being autonomous innovation entities. This demonstrates the nature of an integrated technology-industry complex under national strategy, unlike models where private companies are central as independent market actors. Furthermore, data is redefined as a strategic asset directly linked to national security beyond market competition factors, and its collection, distribution, utilization, and cross-border transfer are increasingly brought under state control. Legal, ethical, standardization, intellectual property, and security oversight systems are also proactively established to form a comprehensive oversight structure encompassing everything from algorithm design to application, characterized by a systematic approach to promoting large-scale technology diffusion within controllable limits.[9]

Thanks to these structural characteristics, the state-led offensive strategy possesses strengths such as concentrated resource mobilization, consistent long-term strategy implementation, integrated coordination of technology, industry, and military, and rapid construction of large-scale infrastructure. In particular, the ability of the state to concentrate resources on strategic industries by integrating fiscal, financial, and industrial policies during the technology catch-up phase is advantageous for rapid capability accumulation. However, at the same time, the centralized bureaucratic structure may limit the diversity of radical and disruptive innovations, and the designation of strategic industries and large-scale subsidies carry the risk of resource misallocation and errors in political judgment. If corporate autonomy becomes excessively subordinate to national strategy, flexibility and creativity in global market competition may weaken, and strengthened data control and expanded state intervention can become constraints on international cooperation and trust-building. Furthermore, the increasing integration of technological development with national strategy can intensify the tendency for international political conflicts to spill over into the technological domain, which is also pointed out as a structural limitation.[10] In other words, the state-led offensive strategy is a type of strategy whose sustainability depends on how it secures institutional flexibility and international openness, while possessing strengths in high mobilization capacity and long-term strategic implementation.

3. Market-Shaping Defensive Strategy

The market-shaping defensive strategy recognizes AI as a strategically core technology, but rather than securing an offensive advantage in the international technology hegemony competition, it is a type of strategy that seeks to defend its own values, institutions, and social stability by redesigning the institutional conditions and norms of the market. In this strategy, the state functions not merely as an industrial facilitator but as a norm designer and risk manager. While the driving force for technological innovation still resides in the private sector, the state structurally shapes the path of technological development through legal regulations, standardization systems, data governance, competition policies, and accountability mechanisms. That is, rather than closing markets or having the state directly control production, it is a strategy that sets institutional boundaries to ensure that technological development does not conflict with democracy, human rights, personal data protection, and social trust, by reconstructing the rules of market operation.

The most representative example of this strategy is the EU AI Act adopted by the European Union. This regulation, adopted in 2024, is the world's first comprehensive AI regulatory framework, with the 'risk-based approach' as its core principle. It categorizes AI systems into 'unacceptable risk,' 'high risk,' 'limited risk,' and 'minimal risk' based on their risk level, and differentiates the intensity of regulation in proportion to the degree of risk. Systems such as social credit scoring, manipulation of vulnerable groups, and extensive real-time remote biometric identification are prohibited, and high-risk AI used in areas such as education, employment, credit, law enforcement, immigration, and justice must meet strict requirements for risk management, data quality assurance, technical documentation, human oversight, accuracy, robustness, and cybersecurity. Furthermore, for General Purpose AI, models with computational scales above certain thresholds are subject to additional assessment and reporting obligations to manage systemic risks proactively. The EU's approach is also characterized by being part of a comprehensive market-shaping strategy that reconstructs the overall digital order, rather than being an isolated AI regulation. The strengthening of personal data protection through the General Data Protection Regulation (GDPR), the regulation of platform market dominance through the Digital Markets Act (DMA), and the enhancement of online platform accountability through the Digital Services Act (DSA) combine with AI regulation to form a coherent normative framework.[11] Through this, the EU is pursuing a strategy of 'normative power,' aiming to shape the operating principles of the global market through norms and standards, rather than directly securing industrial superiority in the technology hegemony competition.

The strengths of the market-shaping defensive strategy are clear. First, by institutionalizing the social and political risks that technological development may entail, it can secure long-term stability and democratic legitimacy. Second, a clear regulatory framework provides predictability for companies and investors and enhances consumer trust. Third, internationally, by preempting norms and standards, it can exert influence in the process of shaping the global technological order. This can be evaluated as an attempt to secure strategic autonomy in terms of norm competition, beyond simple industrial competition. However, this strategy also entails structural limitations. The most critical issue is the risk of 'proactive over-regulation' that arises from introducing comprehensive regulations before the technology matures. Strict ex-ante regulation increases compliance costs and places a particularly heavy burden on startups and SMEs with limited capital and legal capacity. This can hinder innovative experimentation and rapid commercialization, weakening the dynamism of the industrial ecosystem.

Furthermore, the complex regulatory system may paradoxically favor large corporations. Global companies with sufficient resources can absorb the costs of regulatory compliance, but new companies may have incentives to abandon market entry or relocate outside the EU. Moreover, while the EU has secured a leading position in norm-setting capabilities, its industrial base is relatively weak compared to the US and China in terms of developing super-large models and large-scale computing infrastructure.[12] In this situation, if strong regulations are introduced early, the structural imbalance where 'rule-making power' is exercised but 'technology-producing power' is limited may be exacerbated. This could, in the long run, create tension with the goal of securing strategic autonomy.

The market-shaping defensive strategy is an approach that seeks to defend social values and institutional stability by structurally shaping the direction of technological development through laws and institutions, without fully controlling the market. The EU AI Act is the most systematic implementation of this strategy, representing an attempt to redefine technology competition not merely as industrial hegemony competition but as norm competition. However, this strategy requires continuous balancing between normative legitimacy and industrial competitiveness, and its long-term performance will be determined by how this balance is designed.

4. State-Led Defensive Strategy

The state-led defensive strategy recognizes AI as a core strategic technology for national security and economic development, but it is a type of strategy that focuses on protecting and securing the self-reliance of the domestic industrial base, led by the state, rather than directly pursuing offensive expansion in the global technology hegemony competition. In this strategy, the state functions as the leader in fostering industry and the enforcer of market protection. Rather than relying entirely on autonomous market competition, the government mobilizes various policy tools such as subsidies, public procurement, R&D support, regulation of foreign companies, data control, and infrastructure investment to protect the domestic AI ecosystem from external competitive pressure. This can be seen as an attempt to prevent industrial dependence that may arise from global leading companies monopolizing technology, data, and platforms, and to institutionally create a 'strategic buffer space' for the growth of domestic industries.

This defensive strategy is evolving beyond simple market protection by combining with the concept of Sovereign AI. Sovereign AI refers to a strategic approach that seeks to autonomously build not only data sovereignty, which manages data according to domestic legal frameworks, but also the entire AI model design, training, deployment, and operation, as well as the supporting computing infrastructure and governance systems. This extends beyond the question of 'who controls the data' to the more fundamental issue of sovereignty: 'who designs the algorithms and models, and according to what rules and values is AI operated?' The state-led defensive strategy can be seen as an approach to achieve long-term technological internalization and strategic autonomy by securing such AI sovereignty.[13]

India can be classified as a prime example of this state-led defensive strategy. While recognizing AI as a key driver of economic growth and digital transformation, India has been wary of its domestic AI ecosystem becoming structurally dependent on the platform companies of the US and China. Against this backdrop, NITI Aayog presented a national strategy vision by releasing the "National Strategy for Artificial Intelligence: AI for All" in 2018. Although this strategy superficially emphasizes inclusive growth, its content clearly demonstrates the characteristics of a state-led defensive strategy aimed at securing technological self-reliance and sovereignty.[14]

The AI for All strategy in India is designed so that the state leads the direction of AI development through demand creation centered on the public sector. India has selected priority application areas such as health, agriculture, education, smart cities, and smart mobility, which have significant social externalities. This approach prioritizes social optimization over short-term profit maximization and implies an intention to prevent the AI industry from being reorganized into a commercial consumption market centered on foreign platform companies. In particular, by creating demand directly in areas with low private investment incentives, such as agriculture and health, the state aims to guide the direction of industrial development towards public objectives.

Second, this strategy aims to systematically organize the research ecosystem at the national level through a dual structure: CORE (Centres of Research Excellence) for strengthening basic research capabilities and ICTAI (International Centres for Transformational AI) focused on applied and industrial applications. This is not simply about adopting foreign technology but is a phased strategy to internalize technology in the long term through adaptation and application. It can be evaluated as a choice to leverage the 'late-mover advantage' to accumulate unique capabilities, while acknowledging the constraints of being a latecomer.

Third, it emphasizes securing data sovereignty and platform control. India seeks to maintain data control domestically through requirements for data localization, streamlining of personal data protection systems, and establishment of sector-specific regulatory frameworks. Furthermore, through the concept of the National AI Marketplace (NAIM), it is designing a public platform that integrates data collection, annotation, and model deployment, thereby mitigating structural dependence on global big tech platforms and enabling the state to function as a coordinator and designer of the ecosystem. This can be interpreted as an attempt to build the institutional foundation for Sovereign AI.

Fourth, it concurrently pursues human resource and institutional framework development. Large-scale retraining programs, decentralized education models, and strategies for creating new jobs such as data annotation can be understood as defensive measures to mitigate the impact of automation and minimize social conflict. At the same time, improvements in intellectual property (IP) systems that reflect the specificities of AI patents and data-driven models are institutional preparations to avoid dependence on an IP structure centered on foreign companies.

The strengths of this state-led-defensive strategy include the following. First, it can reduce external technological dependence and build a foundation for long-term industrial self-reliance. Protective policies prevent domestic firms in their nascent stages from being eliminated in direct competition with global giants and secure time for technological accumulation. Furthermore, the Sovereign AI approach, which aims to keep data and infrastructure under domestic control, contributes to mitigating security risks and maintaining policy autonomy. Moreover, an AI development model centered on public objectives can secure social legitimacy and help alleviate issues of inequality and exclusion that may arise during the digital transformation process. However, the state-led-defensive strategy also has structural limitations. Excessive protectionism risks weakening competitive pressure, thereby diminishing the efficiency and innovative capacity of domestic industries. While strengthening regulations on foreign firms may offer short-term market defense effects, it also carries the potential for reduced foreign investment and diminished technological cooperation in the long run. Additionally, building Sovereign AI entails enormous costs, including securing large-scale computing infrastructure, energy resources, and skilled personnel, and developing countries face technological constraints in establishing their own models. Furthermore, state-led resource allocation carries the risk of being accompanied by political judgment and bureaucratic inefficiency.

The state-led-defensive strategy can be considered a realistic response option for latecomer countries in the global AI competition, characterized as a defensive, sovereignty-centric strategy aimed at preventing structural dependency on external technological powers and securing technological sovereignty. India's strategy, under the discourse of inclusive growth, aims to build Sovereign AI, demonstrating an approach that prioritizes ecosystem control and technological internalization over market liberalization. This can be evaluated as a typical example of a state-led AI development strategy that prioritizes technological survival and strategic autonomy, distinguishing it from an offensive hegemonic strategy.

5. Diversity of National AI Strategies

The four national AI strategies examined thus far can be categorized based on their approach to global technological competition and their definition of the relationship between the state and the market. The former concerns the strategic orientation—whether to expand influence in the international order through AI or to manage dependence and risks—while the latter concerns the identity of the actors driving innovation.

The market-led-offensive strategy pursues global AI hegemony by leveraging the innovative capacity and capital accumulation capabilities of private enterprises as its core driving force. In this strategy, the state acts as a facilitator, creating an innovation-friendly environment through deregulation, R&D support, talent attraction, and tax incentives, rather than directly controlling industries. The state-led-offensive strategy aims to maintain offensive objectives but achieves them through state-centric industrial policies and resource mobilization. In this case, the state emerges not merely as a coordinator but as a key actor that designs industrial direction and allocates resources to strategic sectors. The market-adjusted-defensive strategy seeks to maintain the basic structure of a market economy while managing the social risks posed by AI technology through robust regulation and institutional design. This type prioritizes the protection of values such as safety, ethics, human rights, and data protection over offensive expansion in the global hegemonic competition. The state-led-defensive strategy involves the state leading industrial development, but its objective is to prevent technological dependence and secure strategic autonomy rather than global hegemony. This strategy pursues the internalization of its domestic AI ecosystem through industrial protection, data control, public demand creation, and infrastructure self-reliance.

Table 3: Diversity of National AI Strategies

These four types of national AI strategies represent strategic combinations chosen based on a nation's capabilities, industrial structure, political system, and international standing. They are not mutually exclusive and may appear in mixed forms depending on the structural conditions each country faces. However, this typology provides a useful analytical framework for understanding the structure of global AI competition and for comparatively analyzing the implications of each nation's strategic choices.

III. Interaction of Global AI Competition and National Strategies

While AI development strategies originate from individual national policy choices, they do not operate in isolation. As a General Purpose Technology, AI has ripple effects across economic, industrial, military, and social domains, and its pace of development and scope of diffusion possess the potential to reshape national competitiveness and the hierarchical structure of the international order. Furthermore, the accumulation of large-scale data, securing of computing resources, cumulative learning effects of algorithm performance, network effects, and platform ecosystem dependencies create structures that allow leaders to widen the gap more rapidly. In such structures, winner-take-all or winner-takes-most phenomena are likely to occur. Nations or companies that first capture the market absorb additional data, capital, and talent, forming a self-reinforcing advantage, while latecomers require exponentially greater costs and time to reach the same level. Thus, the AI industry has a competitive structure with very high delay costs, and the perception that 'once you fall behind, it is difficult to catch up' exerts strong pressure on national strategic choices.

Amidst these competitive pressures, national AI strategies tend to converge through interaction. In particular, as AI hegemonic competition intensifies, there is a growing likelihood of convergence between the US's market-led-offensive strategy and China's state-led-offensive strategy, which could promote technological bloc formation and supply chain fragmentation. Furthermore, the race for technological speed driven by competitive pressure makes the realization of market-adjusted-defensive strategies difficult, increasing the risk of failing to adequately manage social externalities such as labor market shocks, information asymmetry, and widening inequality. The indiscriminate pursuit of Sovereign AI strategies can also lead to resource waste and strategic inefficiency. Thus, as AI competition and technological diffusion accelerate, risks increase in terms of global economic and social stability and the efficient utilization of national strategic resources, posing challenges that cannot be managed by national-level responses alone. Therefore, establishing international AI governance is required.

1. Global AI Competition

On September 1, 2017, Russian President Vladimir Putin stated in a public lecture with students, "Whoever becomes the leader in this sphere will be the ruler of the world."[15] His remarks, beyond mere rhetoric, encapsulated the perception of AI as a strategic technology that would determine the direction of national power and the international order. Today, AI is considered not just a tool for industrial innovation but a technology with extensive influence over military power, information warfare, economic productivity, and norm-setting capabilities, leading to an increasingly zero-sum nature in global competition.

Most importantly, the key factor defining the nature of AI competition is that AI is a general-purpose technology. Historically, general-purpose technologies such as the steam engine, electricity, and information and communication technology (ICT) have fundamentally reshaped production methods and organizational structures across entire economies, beyond specific industries. AI can be applied in almost all fields, including manufacturing, finance, defense, healthcare, education, and administration, and possesses the potential to completely transform the efficiency and structure of existing industries through data processing, decision-making, and automation. Due to this generality, the productivity improvement effects may initially be limited, but if organizational structures and systems are reorganized accordingly, its ripple effects can expand exponentially.[16] This general nature of AI universalizes the scope of competition. It triggers not just competition between firms within a specific industry but a 'total competition' encompassing a nation's entire industrial ecosystem, data infrastructure, talent development system, military strategy, and international standard-setting capabilities. In other words, AI competition is an R&D competition, a capital mobilization competition, a talent acquisition competition, and a competition over platforms and norms. Within this multi-layered competitive structure, governments and corporations perceive AI development not merely as an economic opportunity but as a means to secure strategic advantage that can reshape the foundation of national competitiveness.

Furthermore, the AI industry exhibits strong structural characteristics of winner-take-all or winner-takes-most. The network effects and data accumulation effects observed in the digital platform industry are further amplified in the AI sector. Algorithms improve their performance with more data, and as performance improves, they form a self-reinforcing structure that absorbs more users and data.[17] Once a company or nation secures a leading position, it can widen the gap by attracting additional capital, talent, and ecosystems. Conversely, latecomers must invest enormous costs and time to reach the same level, increasing the risk of falling into technological and market dependence. This structure creates strong incentives to accelerate AI development rather than suppress it, ultimately intensifying competition.

Ultimately, global AI competition, driven by the structural characteristics of a general-purpose technology and a winner-take-all market structure, takes on the nature of an all-out war, pressuring both nations and corporations with the imperative 'not to fall behind.' While this has strategic implications comparable to the nuclear arms race during the Cold War, its influence extends far beyond the military domain into the economy and society as a whole, making it even more pervasive. Therefore, AI competition is not merely an issue of technological innovation but is evolving into a comprehensive power struggle over speed, scale, ecosystems, and norms.[18]

2. Convergence of National Strategies Amidst AI Hegemonic Competition

As global AI competition intensifies, national AI development strategies are showing a tendency to converge through mutual checks, imitation, and institutional learning. In particular, as the intensity of AI hegemonic competition between the US and China, which have adopted offensive strategies, increases, the emphasis shifts towards securing short-term advantages and strategic control rather than long-term ecosystem building or open cooperation.

First, the convergence between market-led-offensive and state-led-offensive strategies is clearly evident in the US-China competition. Traditionally, the market-led-offensive strategy focused on securing technological superiority based on the innovative capacity of private enterprises and an open market order. In contrast, the state-led-offensive strategy involved concentrated development of strategic industries through large-scale state investment and industrial policies. However, as AI competition is increasingly perceived as a competition in strategic industries directly linked to national security, both strategies are gradually beginning to employ similar policy instruments. Notably, the US, which currently holds a relative advantage in AI, is increasingly strengthening its state-led and market-protective characteristics through semiconductor export controls, restrictions on investment in advanced technologies, and large-scale public subsidy policies. This can be interpreted not merely as technological protection but as a strategy to delay the catch-up of competing nations and structurally solidify its own technological dominance. Consequently, the market-led-offensive strategy is gradually converging with the state-led-offensive strategy, showing a tendency to move towards the formation of bloc-based technological spheres rather than an open global economic order.

Furthermore, this convergence is not limited to increased external protectionism but also manifests internally as intensified centralization and industrial concentration. Generally, the market-led-offensive strategy focused on securing technological superiority based on the innovative capacity of private enterprises and an open market order, viewing a decentralized ecosystem with numerous competing companies and startups as the source of innovation. However, as the understanding that AI performance is enhanced by the 'scale' of massive data and vast computing resources spreads, the logic that it is more efficient for AI competition to be led by a few 'national champions' or large platform companies is gaining traction. However, such centralization carries the risk of weakening the engine of innovation in the long run. Innovation is typically fostered through diverse experiments, reduced entry barriers, and the free movement of talent and entrepreneurship, but excessive concentration can restrict this competitive dynamism.[19]

Consequently, US-China competition is converging strategies towards increased state intervention and industrial concentration to secure short-term competitive advantages. While this may contribute to widening the technological gap and strengthening strategic control in the short term, it carries the potential to limit productivity growth by reducing innovation diversity and experimental scope in the long run. Simultaneously, this centralization and bloc formation can accelerate the fragmentation of global supply chains and the digital economy, leading to structural effects that shrink the open global economic order. AI competition is thus converging national strategies in a direction that strengthens protectionism and technological bloc formation externally, and industrial and central concentration internally.

3. AI Competition and Socioeconomic Stability

The socioeconomic effects of AI technology diffusion have not yet been clearly verified by sufficient and timely data. However, considering that during the early stages of the Industrial Revolution, real wages stagnated for a long period despite increased output per worker, and it took considerable time for technological progress to translate into worker welfare, it can be assessed that AI, while possessing the potential to enhance productivity and growth rates, is likely to reduce labor demand and worsen income distribution in the short term through the 'displacement effect.' In particular, whereas previous robotic automation primarily replaced 'low-education, low-skill, low-wage' labor, AI has the potential to automate a significant portion of tasks performed by 'high-education, high-skill, high-wage' professionals. This suggests that the diffusion of AI could also significantly impact professionals and office workers in developed countries, and the possibility of a decline in the labor income share and an widening gap between wages and productivity cannot be ruled out. Of course, technology can also restore labor demand through the 'reinstatement effect,' creating new tasks. Throughout the 19th and 20th centuries, the advent of the steam engine, electricity, and computers created new occupations that did not exist before, which in the long run drove increases in wages and living standards. However, some scholars point out that recent ICT and AI innovations have focused on automation rather than the creation of new tasks, resulting in stagnant labor demand and widening inequality.[20]

The market-adjusted-defensive strategy can be viewed as an attempt to manage these distributional issues and social shocks in advance. To manage labor market shocks, widening inequality, and other issues that may arise from the rapid diffusion of AI, establishing norms and supervisory frameworks is essential. However, under intense AI competition pressure, the market-adjusted-defensive strategy faces a structural dilemma. This strategy prioritizes risk control, norm setting, and securing social trust, based on an approach that preemptively manages externalities arising from technological diffusion. However, in an AI industry with strong winner-take-all characteristics and economies of scale, where AI performance improves non-linearly with data, computing resources, and network effects, initial speed differences are likely to become entrenched long-term technological gaps. In this context, the market-adjusted-defensive strategy, which invests time in establishing regulations and safeguards, may impede innovation speed in the short term and place it at a disadvantage in the race for technological leadership. This is why maintaining normative consistency becomes increasingly difficult as competitive pressure intensifies.

Recently, the need to secure AI competitiveness has been increasingly emphasized in Europe as well. Facing pressure on economic growth and competitiveness due to an aging population and slowing productivity, the EU is at risk of falling behind in the global AI competition. To overcome this, there is a growing call for decisive action and ambitious goal-setting, primarily by private companies. In other words, the demand for the EU to also focus on securing AI competitiveness through cooperation in technology, industry, and policy, mobilization of talent and capital, modernization of key infrastructure, and creation of a competitive technological ecosystem is gradually spreading.[21]

Ultimately, the market-adjusted-defensive strategy entails a highly complex choice between mitigating the risk of widening technological gaps and the risk of undermining social stability. As competitive pressure intensifies, the temptation to ease regulations and support industries grows, but this can weaken social safety nets. Conversely, maintaining a strict regulatory stance increases the risks of technological dependence and slowed growth. This dilemma suggests that as AI competition intensifies, it becomes more difficult to implement measures for managing social risks. Therefore, the measures that must be taken proactively to manage the social risks posed by AI are to mitigate AI competition itself.

4. AI Competition and Sovereign AI

AI competition is also acting as a key factor in promoting the spread of 'Sovereign AI' strategies. As competition intensifies, dependence on external platforms and technologies is perceived as a security vulnerability, leading each nation to strengthen policies aimed at securing autonomy, such as data localization, building domestic cloud infrastructure, and developing indigenous AI models. This trend is observed not only in state-led defensive strategies but also in countries pursuing offensive strategies. In other words, AI competition is converging national policies towards valuing technological sovereignty, regardless of the strategy type.

While AI competition is thus promoting Sovereign AI strategies, there are several structural constraints and limitations to the feasibility and practicality of such strategies. First, Sovereign AI presupposes a strategic approach that aims to place extensive resources, including core technologies, data, computing infrastructure, standards, and talent, under national control. However, the core areas of the AI industry are already converging into a global oligopoly or monopoly structure. In a situation where specific companies and nations hold market dominance, it requires immense cost and time for latecomer countries to achieve complete self-reliance in these core domains. For example, the reality that Microsoft-OpenAI dominates about 70% of the commercial LLM market and a few companies supply over 90% of GPUs for AI training makes it difficult for nations to solely control these core resources and technologies.

Second, the practical feasibility of Sovereign AI strategies heavily depends on the size of the domestic market and the capacity of the industrial ecosystem. Countries with large domestic markets can achieve a certain level of technological and industrial internalization through protectionist approaches, but it is difficult for countries with limited domestic markets to implement such strategies. A limited market size poses constraints in terms of recouping R&D investments, securing talent, and fostering experimental innovation. Protectionist policies may paradoxically hinder innovation and increase costs.

Consequently, despite the logical necessity of Sovereign AI strategies to secure technological sovereignty and national security amidst competitive pressure, they are not strategies that all countries can equally implement due to the realistic constraints of global industrial structure, domestic market size, technological concentration, and access to capital and talent. Therefore, even as AI competition intensifies, each country must prioritize strategic objectives and adopt an approach of pursuing selective Sovereign AI in core areas and specialized markets. A hybrid strategy of global cooperation and internalization, rather than complete self-reliance, may be a more realistic and effective alternative.

5. AI Competition and Governance

AI competition provides strong incentives for both nations and corporations to secure technological superiority and strategic autonomy, but it can also cause structural problems for the global economy, society, and national strategies overall. First, as competition intensifies excessively, convergence between market-led offensive strategies and state-led offensive strategies is occurring, and there is a high probability that technological bloc formation and supply chain fragmentation will be promoted, centered around the US and China, rather than an open global economy. This can undermine the efficiency of global trade and the digital economy, and in the long term, weaken the openness of the world economy.

Second, the race for technological speed driven by competitive pressure makes it difficult to realize market-adjustment defensive strategies, increasing the risk of inadequately managing social externalities such as labor market shocks, information asymmetry, and deepening inequality. With the winner-take-all nature and rapid diffusion of AI, defensive strategies that require time to establish regulations and safety nets may limit the speed of innovation in the short term while failing to adequately ensure social stability.

Third, the indiscriminate pursuit of Sovereign AI strategies can also lead to resource waste and strategic inefficiency. Overly protectionist policies aimed at self-sufficiency in core technologies and infrastructure require substantial investment and talent mobilization, but given the industrial structure, domestic market size, and potential for technological dependence, it is difficult for all countries to implement them successfully. This carries the risk of increasing the cost burden and limiting innovation capacity in some countries.

As AI competition and technological diffusion accelerate, risks increase in terms of global economic and social stability, and the efficient utilization of national strategic resources. These are problems that cannot be managed by individual countries alone. Therefore, establishing international AI Governance is becoming an essential task to prevent the side effects and unintended consequences of new technologies.[22]Kissinger and Allison emphasized, "The possibility that the unlimited development of AI could have catastrophic consequences for the United States and the world is too great for government leaders not to act immediately," clarifying that AI is not merely an economic or technological issue but a potential threat that could determine global strategic stability and the future of humanity.[23]Furthermore, AI developers have warned that "the existential risk posed by AI should be treated as a global priority on par with societal-scale risks such as pandemics and nuclear war," recalling Turing's concern that AI could control human lives. They stressed the need for a robust oversight system and new regulatory frameworks that ensure ethical, transparent, and controllable innovation to manage the fundamental changes (GenAI) that AI will bring.[24]

However, despite these warnings, the establishment of AI Governance is considered one of the most challenging tasks facing the international community.[25] AI Governance should be established not merely as a technical regulation, but as a comprehensive, multi-layered system to mitigate the weakening of global economic openness, the infringement of social stability, and the resource waste resulting from the indiscriminate pursuit of Sovereign AI strategies due to AI competition and technological diffusion. This implies that AI policy is not confined to technology and industry but involves considering complex factors such as international cooperation, standard coordination, and ethical and social consensus. However, given that AI is a general-purpose technology (GPT) with a winner-take-all structure, and national strategies are intertwined, establishing governance is inevitably an extremely difficult task.

While establishing AI Governance is a very difficult task, the case of the Basel Accord can serve as a useful reference. The Basel Accord emerged from the need for major countries, including the United States, to strengthen domestic financial regulations to ensure financial market stability following the banking crises of the 1970s, while simultaneously addressing the issue that strengthening regulations could undermine the international competitiveness of domestic financial institutions. In other words, it can be seen as a case where international norms were used to resolve the problem of domestic regulations potentially weakening competitiveness. Through the Basel Accord, the United States secured the stability of its domestic financial market while also inducing other countries to apply the same standards, thereby balancing the competitive environment.[26]

The logic that emerged during the establishment of the Basel Accord can be similarly applied to AI Governance. That is, while introducing strong regulations unilaterally for financial market stability carries the risk of weakening competitiveness, the fact that balance between stability and competitiveness could be achieved through international standards and norms offers implications for the AI field as well. The rapid diffusion of AI heightens concerns about social stability, including labor market shocks, information distortion, and deepening inequality, which can create pressure to strengthen regulations and safety nets. In particular, if socioeconomic instability caused by AI diffusion intensifies within the United States, which is leading AI technological development, this situation could provide strong momentum for establishing AI Governance. Currently, one such possibility is the AI Bubble problem. There are ongoing arguments that a financial and industrial bubble has formed due to the combination of excessive investment and expectations in the AI industry.[27] If this bubble bursts, voices calling for regulating AI development to some extent and ensuring stability within the United States will likely strengthen, which could also present an opportunity for establishing AI Governance.

IV. Conclusion

The South Korean government currently considers the AI great transformation as a core strategy for national economic growth, aiming to transition the Korean economy from a follower to a leader by becoming one of the top three AI powers. To this end, the government is pursuing comprehensive policy initiatives, including expanding the AI budget, forming an 'AI Highway' by building a national data center, securing high-performance GPUs and AI data cluster hubs, revitalizing convergence industries through the 'AI for All' project and regulatory exceptions, and fostering future talent. These strategies fundamentally exhibit characteristics of a state-led development strategy, while also being designed to leverage private-led innovation, indicating that South Korea is setting its strategic direction along the extension of its existing developmental state model.

Considering the global AI competition landscape, South Korea's approach should involve participating in AI alliances and global cooperation based on its semiconductor industry and digital infrastructure, positioning itself as a linchpin state within the AI ecosystem, rather than pursuing independent development. Furthermore, and more importantly, while the current South Korean AI strategy is proactive in technological competition and industrial development, it is relatively weak in managing socioeconomic risks. Despite AI diffusion potentially causing various social problems such as automation, labor market restructuring, and deepening social inequality, institutional mechanisms and policy responses to mitigate these issues are not sufficiently in place. Realistically, strengthening independent regulations amidst global competition could lead to a decline in competitiveness. Therefore, in the short term, policies that offset negative effects, such as strengthening social safety nets and implementing retraining/labor transition programs, are necessary. In the long term, managing social risks collectively through international AI regulations and governance, and participating in the formation of global standards and ethical guidelines, can simultaneously secure technological competitiveness and social stability.

In this regard, South Korea's AI strategy focuses on domestic industry development and technological infrastructure construction, but it can be assessed as somewhat lacking in AI diplomacy and international cooperation strategies. For South Korea to function as a strategic linchpin state in the fierce global AI competition and to proactively participate in the formation of international norms and governance, an AI diplomacy strategy that integrates technological competitiveness, industrial development, and social stability is necessary. In particular, this approach should not only promote AI development but also include the role of strategically managing the pace of development. Polanyi pointed out that "the rate of change is as important as the direction of change itself," emphasizing that the core role of government in the economic sphere is to "moderate the rate of change."[28] This suggests that South Korea's AI strategy needs a balanced approach that not only pursues technological competition and industrial development but also manages the pace of AI development through diplomatic and policy means to mitigate the risks of AI diffusion.■

[1]Government of Canada. 2022. "Government of Canada Launches Second Phase of the Pan-Canadian Artificial Intelligence Strategy." June 22. https://www.canada.ca/en/innovation-science-economic-development/news/2022/06/government-of-canada-launches-second-phase-of-the-pan-canadian-artificial-intelligence-strategy.html.

[2]Radu, Roxana. 2021. "Steering the Governance of Artificial Intelligence: National Strategies in Perspective. " Policy and Society, 40 (2): pp. 178-193.

[3]Dutton, Tim. 2018. "Building an AI World: Report on National and Regional AI Strategies." CIFAR. December 6. https://cifar.ca/cifarnews/2018/12/06/building-an-ai-world-report-on-national-and-regional-ai-strategies/#topskipToContent.

[4]Weber, Max. 1978. Economy and Society. Berkeley: University of California Press. p. 6.

[5]Radu, Roxana. 2021. "Steering the Governance of Artificial Intelligence."Policy and Society, 40 (2): pp. 179-180.

[6]The White House. 2020. “American Artificial Intelligence Initiative: Year One Annual Report.”

[7]The White House. 2025. “Winning the Race: America’s AI Action Plan.”

[8]Khanal, Shaleen, Hongzhou Zhang, and Araz Taeihagh. 2025. "Why and How Is the Power of Big Tech Increasing in the Policy Process? The Case of Generative AI." Policy and Society, 44 (1): pp. 52–69.

[9]State Council of China. 2017. "The New Generation Artificial Intelligence Development Plan." https://digichina.stanford.edu/work/full-translation-chinas-new-generation-artificial-intelligence-development-plan-2017/.

[10]Frey, Carl Benedikt. 2025. "How the Battle for Control Could Crush AI’s Promise." Finance & Development, 62 (3): pp. 50-53.

[11]EU Artificial Intelligence Act. 2024. "High-level Summary of the AI Act." February 27. https://artificialintelligenceact.eu/high-level-summary/; European Union. 2024. "Regulation (EU) 2024/1689 of the European Parliament and of the Council." June 13. https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689.

[12]Rajan, Raghuram R. 2025. "The Tradeoffs of AI Regulation." Project Syndicate. August 26. https://www.project-syndicate.org/commentary/ai-regulation-innovation-tradeoff-us-versus-europe-by-raghuram-g-rajan-2025-08.

[13]Letort, Brian and Kadri Linask-Goode. 2025. "What Is Sovereign AI and Why Is It Growing in Importance?" Digital Reality. April 3. https://www.digitalrealty.com/resources/blog/what-is-sovereign-ai.

[14]NITI Aayog. 2018. National Strategy for Artificial Intelligence: AI for All.

[15]RT. 2017. "‘Whoever leads in AI will rule the world’: Putin to Russian children on Knowledge Day." September 1. https://www.rt.com/news/401731-ai-rule-world-putin/.

[16]Kishtainy, Niall. 2025. "A New Industrial Revolution?" Finance & Development, 62 (4): pp. 46-49.

[17]Radu, Roxana. 2021. "Steering the Governance of Artificial Intelligence." Policy and Society, 40 (2): p. 189. Johnson, Simon. 2025. "Tech’s Winner-Take-All Trap." Finance & Development, 62 (2): pp. 66-67.

[18]Bremmer, Ian and Mustafa Suleyman. 2023. "Building Blocks for AI Governance." Finance & Development, 60 (4): pp. 10-12.

[19]Frey, Carl Benedikt. 2025. "How the Battle for Control Could Crush AI’s Promise." Finance & Development, 62 (3).

[20]Comunale, Mariarosaria and Andrea Maneara. 2024. "The Economic Impacts and the Regulation of AI: A Review of the Academic Literature and Policy Actions." International Monetary Fund Working Paper, WP/24/65; Kishtainy, Niall. 2025. "A New Industrial Revolution?" Finance & Development, 62, 4: pp. 46-49.

[21]General Catalyst. 2025. An Ambitious Agenda for European AI.

[22]Radu, Roxana. 2021. "Steering the Governance of Artificial Intelligence." Policy and Society, 40 (2): p. 180.

[23]Kissinger, Henry A. and Graham Allison. 2023. "The Path to AI Arms Control."Foreign Affairs. October 13. https://www.foreignaffairs.com/united-states/henry-kissinger-path-artificial-intelligence-arms-control.

[24]Tourpe, Hervé. 2025. “Artificial Intelligence’s Promise and Peril.” Finance & Development 60(4): pp. 8-9.

[25]Bremmer, Ian and Mustafa Suleyman. 2023. "Building Blocks for AI Governance."Finance & Development, 60 (4).

[26]Kapstein, Ethan Barnaby. 1992. "Between Power and Purpose: Central Bankers and the Politics of Regulatory Convergence."International Organization, 46 (1): pp. 265-287; Oatley, Thomas and Robert Nabors. 1998. "Redistributive Cooperation: Market Failure, Wealth Transfers, and the Basle Accord."International Organization, 52 (1): pp. 35-54.

[27]Stiglitz, Joseph E. 2025. "Trump and the End of American Hegemony."Project Syndicate. December 15. https://www.project-syndicate.org/magazine/trump-end-of-american-hegemony-by-joseph-e-stiglitz-2025-12.

[28]Polanyi, Karl. 1944. The Great Transformation. Boston: Beacon Press, pp. 36-37.

Author: Jaehwan Jeong_Professor, Inha University.

Managed and Edited by: Jaehyun Im_EAI Researcher

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