As AI transforms the broader paradigm of society and the economy and emerges as a core foundation of national competitiveness and national security, the role of the state in responding to AI is also changing fundamentally. Early AI policy largely belonged to the domains of technology development and industrial promotion. The main concerns of governments were how much to invest in AI research and development, how to train specialized talent, and how to support companies in developing and using AI. AI was a new growth industry, and the basic role of the state was to promote technological innovation and industrial growth.
However, as AI rapidly spreads across society, the scope and subject of AI policy are also expanding significantly. As AI begins to affect not only technological development in specific industries but also labor and education, healthcare and finance, media and culture, market competition and public administration, national security and the international order, AI policy is expanding beyond industrial promotion and technology regulation into the design of institutions and order across society.
Above all, as AI is used in decisions that have significant effects on human lives, issues of bias, discrimination and accountability have emerged. As generative AI creates text, images, music and video, questions of copyright, creators’ rights, misinformation and the reliability of information have become central. In the process of training on large-scale data, privacy and data rights have become major concerns. As a small number of global companies control foundation models and computing infrastructure, market concentration and competition issues are also becoming more serious. The policy nature of AI has expanded from technology development and industrial promotion to social risk and the public interest.
Yet, as discussed earlier, recent changes go one step further. As AI becomes a core strategic asset not only for economic growth and industrial competitiveness but also for national security, economic security and global power competition, AI policy is expanding into a new dimension. This shift raises new policy questions that go beyond how well a country can develop AI. It asks how effectively AI can be diffused throughout society, and how the risks and conflicts that arise in that process should be managed.
If society cannot trust AI, or if harms and conflicts caused by AI continue to occur, sustainable diffusion will be difficult even for technically excellent AI. Conversely, if excessively strong regulation is applied in an attempt to eliminate all potential risks in advance, corporate innovation, investment and the emergence of new services may be constrained. Therefore, the capabilities that a state must possess in the AI era can be understood in three dimensions: AI technology capacity, AI diffusion capacity and AI governance capacity.
First, AI technology capacity is the ability to develop and continuously innovate AI through AI models, semiconductors, computing, data and talent. Second, AI diffusion capacity is the ability to spread developed or adopted AI across the economy and society, including manufacturing, finance, healthcare, education, media and public administration, and connect it to real productivity and innovation. Third, AI governance capacity is the ability to promote AI innovation while protecting safety, fundamental rights, competition and the public interest, and to design an appropriate balance between global openness and national strategic autonomy.
A country that develops AI, a country that uses AI well and a country that manages AI in a trustworthy way are not necessarily the same. Future AI competitiveness is therefore likely to depend not on maximizing only one of these three capabilities, but on how well they are built and connected in balance. From this perspective, the role of the state in AI policy must also be newly defined.
The state is a promoter that encourages AI research and development and industrial innovation. At the same time, it is a regulator that manages market failures and social risks caused by AI. It is also a capacity builder that develops AI infrastructure, talent, education and social adaptability. And it is a strategist that must secure the nation’s technological and industrial position, economic security, national security and AI sovereignty amid global AI power competition. Modern AI policy is therefore evolving into a complex national policy in which promotion, regulation, capacity building and strategy operate together.
This transformation greatly expands the scope of AI policy itself. AI research and development is science and technology policy. AI semiconductors and data centers are industrial policy and are also connected to energy policy. AI’s impact on labor markets is a matter of labor and education policy. The training and outputs of generative AI fall under copyright and cultural policy. The market power of AI companies and platforms is a matter of competition policy. AI’s influence on information and public opinion formation concerns media and communications policy as well as democracy. The overseas transfer of advanced semiconductors and AI models is both trade policy and economic security policy. The military use of AI belongs to defense, diplomacy and security policy. Then where should the boundary of AI policy be drawn?
This question is one of the important reasons AI policy must be treated as an independent field of study. AI policy is difficult to understand as simply adding one new area to existing policy domains. Rather, it should be understood as a cross-cutting policy domain that reconnects science and technology, industry, competition, labor, education, media, culture, security and diplomacy around the new paradigm of AI. This cross-cutting nature expands AI policy beyond government-centered regulation into a matter of governance involving diverse actors and policy instruments.
AI governance does not mean only that the government makes laws to control companies and technology. It refers to a broader institutional order in which government, companies, developers, users, civil society, academia and international organizations together shape the principles, norms, technical standards and accountability systems governing the development and use of AI. Moreover, as noted earlier, in AI the pace and direction of technological development itself remain uncertain. If detailed rules are made based only on today’s technological level, those rules can quickly become outdated as technology changes. Conversely, if policymakers try to anticipate and regulate every possible future risk in advance, excessive regulation may hinder innovation.
Therefore, the greater the uncertainty surrounding AI becomes, the more difficult it is to predetermine every future situation through detailed rules. This is precisely why basic principles become more important as standards for policy judgment even under changing conditions. It is impossible to predict every AI technology and service that will emerge in the future and turn them into rules today. But it is possible to establish basic principles about what should be protected when new problems arise, which values should be prioritized, and what kind of balance should be sought when values conflict.
In this sense, AI policy is both an entirely new policy domain and one connected to long-standing policy traditions. Principles such as freedom of expression, the public interest, the marketplace of ideas, diversity, competition and universal service have provided core values that society should protect amid rapid technological and market change in the eras of broadcasting, telecommunications and the internet. In the AI era, these principles must be reinterpreted in light of new technological and social realities. At the same time, AI sovereignty, encompassing data, language, culture and technological autonomy, is emerging as a new policy principle.
Ultimately, the core of AI policy is not to chase rapidly developing technologies and regulate individual problems one by one. It is to determine which values should be protected in the new social order created by AI, and what balance and priorities should be designed among values that conflict with one another. At the same time, AI technology and industry must be developed and diffused throughout society, while the risks and conflicts that arise in that process are managed and the nation’s strategic autonomy is secured amid global AI power competition.
In this sense, AI policy must be reestablished as a comprehensive national strategy that designs the future of the state, going beyond the narrower dimensions of regulatory policy or industrial policy.
Therefore, the central policy question in the age of AI transformation is no longer simply, “How should AI be regulated?” or “How much should the AI industry be promoted?”
The question is this:
“What kind of AI society will we build, and according to what principles?”
That is the starting point of AI policy.
