The Multilayered Nature of AI Policy and AI as a Cross-Cutting Policy Problem

1. What Kind of Policy Area Is AI Policy?

As artificial intelligence has emerged as a major national policy agenda, the first question that arises is surprisingly simple: what kind of policy is AI policy?

Is it industrial policy, science and technology policy, or regulatory policy?

Because AI deals with data and personal information, it can be understood as information and communications policy as well as privacy policy. Because it raises issues of algorithmic discrimination and bias, it also has the character of human rights policy. Because AI changes the structure of labor and occupations, it is labor policy. Because it affects education and talent development, it is education policy. And as advanced AI models and semiconductors become strategic assets in national security and the international order, AI policy is expanding beyond industrial and trade policy into the domains of diplomacy and security.

In this sense, AI policy is difficult to classify as one traditional policy domain. It is more appropriate to understand it as a cross-cutting policy problem that traverses existing policy domains and administrative jurisdictions. In policy studies, cross-cutting policy problems refer to issues that cannot be addressed by a single ministry or conventional policy domain alone, and that require the coordination of goals and instruments across multiple policy areas.

Climate change and innovation policy are representative examples. Climate change is not merely an environmental policy issue. It simultaneously touches energy, industry, transportation, agriculture, finance, and diplomacy, making it difficult to address through the policies of individual ministries alone (Candel & Biesbroek, 2016). Innovation policy has also been studied as a representative cross-cutting policy problem because it traverses existing domains such as science, research, technology, and industry, thereby requiring inter-ministerial coordination and policy integration (Christensen & Serrano Velarde, 2019).

AI policy clearly shows these same characteristics. The policy problems raised by AI cut across science and technology, industry, labor, education, personal information, competition, media, culture, diplomacy, and security. Because of this cross-cutting nature, the central task of AI policy is not merely to respond within each policy domain separately. It is to determine how different policy goals and policy instruments should be coordinated and integrated.

AI Policy Began as Technology Policy

Governments’ early interest in AI focused mainly on technological development and industrial promotion. How much should be invested in AI research and development? How should specialized talent be cultivated? How should corporate innovation be supported? These were the main policy questions.

From this perspective, AI was recognized as a strategic technology that the state should foster, much like semiconductors, mobile communications, and biotechnology.

Accordingly, the core instruments of early AI policy were those traditionally associated with science, technology, and industrial policy: R&D support, talent development, data infrastructure, computing infrastructure, and startup support. The basic policy goal was also relatively clear: to promote technological innovation and strengthen industrial competitiveness.

This early policy orientation can also be seen in the United States’ national AI strategy. The National Artificial Intelligence Research and Development Strategic Plan, released in 2016 by the National Science and Technology Council, covered ethical, legal, social, and safety issues related to AI, but its central axis remained the federal government’s research and development policy, including long-term AI research investment, public data and research environments, and the development of an AI research workforce (NSTC, 2016).

However, as the scope of AI application has expanded across society, this approach alone has become insufficient to explain AI policy. Algorithms are now used in socially significant decision-making processes such as loan screening by financial institutions, hiring by companies, diagnosis by medical institutions, content recommendation by platforms, assessment by educational institutions, administrative services by governments, and the selection of welfare beneficiaries. Since the emergence of generative AI, artificial intelligence has also begun to intervene deeply in the very process by which humans produce and consume information and knowledge.

As the social impact of technology expands, the policy questions must also change. Governments now have to consider not only “how to develop AI,” but also “how AI should be used,” “who should be responsible for AI decisions,” and “how the risks created by AI should be managed.”

This is the point at which AI policy expands beyond technological development and industrial promotion into the social management and governance of technology.

AI Policy Is Both Industrial Policy and Regulatory Policy

One of the most distinctive characteristics of AI policy is that it must simultaneously perform two different policy functions: promotion and regulation.

From the perspective of promotion, AI policy is an extension of traditional science, technology, and industrial policy. Governments must expand R&D investment, cultivate AI talent, secure data and computing infrastructure, and create an environment in which companies can develop new AI services. In a global AI race directly connected to national competitiveness, this industrial policy function becomes even more important.

At the same time, governments must respond to the risks created by AI. Problems such as privacy violations, algorithmic bias and discrimination, misinformation, copyright infringement, safety risks, labor market disruption, and concentration of market power are difficult to solve through market self-regulation alone. In areas such as healthcare, finance, transportation, national defense, and public administration, where AI errors can cause serious harm to individuals or society, certain rules and oversight mechanisms are unavoidable.

As a result, the boundary between traditional industrial policy and regulatory policy becomes blurred in AI policy. From the perspective of industrial policy, regulation may be seen as a cost that restricts innovation. Conversely, from the perspective of regulatory policy, excessive industrial promotion may risk sacrificing safety and rights protection. But in AI policy, the problem cannot be solved simply by choosing one over the other. The core policy challenge is precisely to promote innovation while managing risk.

This dual character of AI policy is clearly visible in the recent policy choices of the United States and the European Union.

The Trump administration in the United States, upon taking office in 2025, revoked the Biden administration’s AI executive order, which had emphasized safety, trust, and risk management, and shifted the center of policy gravity toward removing regulatory barriers that were seen as hindering AI innovation (The White House, 2025a). The America’s AI Action Plan, released in July 2025, presented three core pillars: accelerating innovation, building AI infrastructure, and leading in international diplomacy and security. It emphasized private-sector-led technological innovation and the strengthening of America’s global AI competitiveness (The White House, 2025b).

By contrast, the European Union established a comprehensive regulatory framework through the AI Act, which entered into force in 2024 and regulates AI according to levels of risk. The law sets as its main objective the management of risks that AI may pose to health, safety, and fundamental rights, while at the same time promoting the human-centered and trustworthy use of AI and supporting innovation (European Union, 2024).

Of course, U.S. AI policy also includes considerations of safety and risk management, and the European Union also treats innovation and industrial competitiveness as important policy goals. Nevertheless, the difference is clear: the United States places relatively greater weight on innovation and industrial competitiveness, while the European Union places greater emphasis on risk management and the protection of fundamental rights.

The tension between innovation and regulation embedded in AI policy is therefore expanding beyond individual national policy choices into a global competition over norms and regulatory systems in the emerging AI order.

AI Policy Is Both Economic Policy and Social Policy

Another characteristic of AI policy is that economic goals and social goals are strongly intertwined.

AI has the character of a general-purpose technology that increases productivity and creates new industries and markets. It automates corporate business processes, accelerates research and development and software development, and enables the emergence of new products and services. In this respect, AI policy is economic policy concerned with growth, productivity, industrial competitiveness, and employment structures.

A similar policy dynamic appeared during the diffusion of the internet. The U.S. Telecommunications Act of 1996 symbolized a policy shift aimed at promoting innovation in the telecommunications and information industries by introducing competition into existing telecommunications markets and reducing regulatory barriers. It was a strong economic policy approach that sought to expand investment and stimulate the growth of new information and communications services and industries through competition and markets.

However, as the internet spread as a core infrastructure of the economy and society, new problems emerged. Significant differences appeared in people’s ability to access and use computers and the internet depending on income, education level, race, and region. These disparities came to be recognized as a new social problem known as the digital divide.

The National Telecommunications and Information Administration under the U.S. Department of Commerce began investigating disparities in access to information and communications technologies through its Falling Through the Net reports in the mid-1990s. In its 1999 report, Defining the Digital Divide, the NTIA pointed out that the digital divide persisted among low-income groups, people with lower levels of education, racial minorities, and rural communities, and that in some areas it was even widening (NTIA, 1999).

In response, the Clinton administration began to treat the digital divide not merely as a matter of technological access, but as a problem of economic and social participation. The 2000 report From Digital Divide to Digital Opportunity promoted expanded internet access for low-income groups and underserved regions, the development of Community Technology Centers, information technology education and job training, and public-private partnerships in order to address gaps that market competition and technological development alone could not solve. The policy target was expanded beyond access to technology itself to include the capacity to use it effectively (The White House, 2000).

The experience of internet policy showed that policy for a new technology cannot remain limited to economic growth and industrial innovation. As technology becomes core social infrastructure, the questions of who can access it, who can use it, and how its benefits are socially distributed become central policy issues. The experience of internet policy, which expanded from an economic policy promoting competition and innovation into a social policy aimed at reducing the digital divide, provides an important precedent for AI policy.

The diffusion of AI raises similar but even more complex problems. A new AI divide may emerge between individuals who can use AI and those who cannot, between companies capable of adopting AI and those that are not, and between countries that possess large-scale computing infrastructure and data and those that do not. Moreover, the divide in the AI era is not simply a matter of whether one can access technology. It may appear as a difference in how effectively one can use AI, what level of AI capability and service one can access, and how far AI use can be linked to productivity, income, education, and employment opportunities.

Therefore, AI policy must address not only technological development and economic performance, but also how the opportunities and risks created by AI are socially distributed. This is why policy issues such as AI access, AI literacy, labor transition, education and retraining, digital inclusion, and public AI are becoming increasingly important. Ensuring that the benefits of AI are not enjoyed only by a small number of companies or expert groups, but can be broadly used by members of society, becomes a new public-interest task in the AI era.

AI Policy Is Both Domestic Policy and International Policy

Another defining feature of AI policy is that the boundary between domestic policy and international policy is rapidly weakening. While AI technology development and industrial promotion take place within individual national policy domains, the core resources and markets that make them possible are already globally connected. Advanced semiconductors, computing infrastructure, cloud services, data, AI models, research talent, and capital move through global supply chains and platforms.

Therefore, a country’s AI capability is not determined by domestic R&D investment alone. It also depends on whether the country can secure advanced AI semiconductors, access global cloud services and AI models, and participate in international research and talent networks.

An even more important change is that AI has emerged as a core strategic asset in U.S.-China strategic competition. The United States has used export controls on advanced AI semiconductors and related technologies to constrain China’s access to the computing capabilities needed for frontier AI development, while also pursuing a strategy of expanding U.S. AI technology and industrial ecosystems into the global market.

The international character of U.S. AI strategy is especially clear in the 2025 America’s AI Action Plan. The plan treats U.S. AI competitiveness not merely as a matter of domestic industrial policy, but as a question of international diplomacy and security. It proposes spreading what it calls the American AI “full stack” — including hardware, AI models, software, applications, and standards — to allies and partner countries (The White House, 2025b).

The U.S. government is pursuing this not merely as a policy idea, but as a concrete external strategy. The July 2025 executive order, Promoting the Export of the American AI Technology Stack, presents the global adoption of U.S. AI technology, standards, and governance models as national policy and establishes the American AI Exports Program. The aim is to export AI hardware, cloud, data systems, models, and applications as an integrated package (The White House, 2025c). This is a representative example of how U.S. AI policy is expanding beyond industrial policy into trade, diplomacy, and security policy.

China, for its part, is responding to U.S. technological controls by strengthening its own AI technology and industrial ecosystem while seeking to expand its influence in international AI cooperation and governance. In its 2025 Global AI Governance Action Plan, China defined AI as an international public good and emphasized respect for national sovereignty, open cooperation, infrastructure development, shared use of technology, and the reduction of the global AI divide (Ministry of Foreign Affairs of the People’s Republic of China, 2025).

The U.S.-China AI competition is therefore expanding beyond a technological race over which country can develop superior models. It is becoming a competition over the global order surrounding semiconductors and computing infrastructure, models and platforms, international standards, and governance norms.

These developments raise particularly important policy questions for countries such as Korea. Even if AI technologies and services are developed domestically, if they depend heavily on advanced GPUs, global cloud infrastructure, and general-purpose AI models and platforms developed overseas, then changes in other countries’ export controls, technology policies, or service provision policies can directly affect the domestic AI industry and AI-use environment.

Restrictions on the supply of AI semiconductors can affect the computing capacity of domestic companies. Changes in the model-provision conditions or service policies of global AI providers can directly influence how businesses and public institutions use AI. In this sense, other countries’ AI policies become variables that shape one’s own AI policy environment.

Therefore, AI policy can no longer be designed only as a policy for fostering domestic industries and protecting domestic users. It must also consider the stable securing of semiconductors and computing infrastructure, dependence on global AI models and platforms, data and technological sovereignty, international standards, export controls, trade rules, alliances, and international cooperation.

At the same time, it must respond to a new policy tension: how to maintain the openness of the global AI ecosystem while securing a certain level of national autonomy over core technologies and infrastructure.

In this respect, AI policy is both domestic policy and international policy, combining diplomacy, trade, and security. It is also strategic policy that determines a country’s technological autonomy and its position in the global AI order.