There is little disagreement that AI is one of the most important advanced technologies of our time. Based on large language models, GPUs and AI semiconductors, data centers, cloud computing, training data and algorithms, AI’s technical capabilities are developing at remarkable speed. The problem is that it is difficult to accurately predict the pace and direction of that development, as well as the level of capability AI may ultimately reach.
The development of generative AI clearly shows this uncertainty. Only a few years ago, it would have been difficult to imagine that AI capable of conversing naturally with humans, writing professional documents, developing software, and generating images and videos would become widespread so quickly. More recently, AI has been moving beyond generative AI, which responds to human questions and creates content, toward agentic AI, which can pursue goals, use necessary tools and execute a series of tasks on its own.
The first difficulty facing AI policy today, therefore, is the fundamental gap between the speed of technological development and the speed of policymaking. Traditional policy is generally created by identifying current problems, analyzing their causes, setting policy goals, and then designing specific instruments and implementation schedules. In the case of AI, however, the technology itself changes rapidly even while policy is being designed. By the time regulations built around today’s level of technology are implemented, new models and services may already have appeared, along with new risks and opportunities.
This characteristic makes it difficult to turn AI policy into a roadmap or strategy based on a concrete timetable. Long-term workforce policy, education policy, industrial policy, regulatory policy and national strategy all require at least some ability to anticipate the level of technology five or ten years from now. In AI, however, that very premise remains uncertain.
Yet social uncertainty is just as important as technological uncertainty. It is difficult to predict exactly how much AI will replace or complement human labor, how much it will improve corporate productivity, and how it will change the composition of existing and new jobs. In education, AI raises fundamental questions about what humans should learn and what capabilities they should develop. In media and culture, the boundary between human and AI creation is changing, and AI is beginning to play a new gatekeeper role in the production and circulation of information. In politics and democracy, new questions are emerging about AI’s impact on opinion formation, political communication, misinformation and elections. In this sense, social uncertainty refers to the difficulty of predicting how AI will transform the broader social system, including the economy, labor, education, culture, media, democracy and even human decision-making itself.
Another factor that makes AI policy even more difficult is that geopolitical uncertainty is now layered on top of technological and social uncertainty. As AI emerges as a strategic asset that affects not only national competitiveness but also economic security and national security, competition over AI is expanding beyond technological competition among companies into competition among states. In particular, as rivalry between the United States and China intensifies over AI technology, infrastructure, supply chains and global norms, countries and economic blocs are beginning to view AI through increasingly different perspectives and policy interests.
On one side, there is the view that AI development and market autonomy should be protected as much as possible in order to promote innovation and industrial competitiveness. On the other, there is the argument that stronger regulation and accountability are needed to respond to risks to fundamental rights and society. As national security and technological sovereignty become connected to this debate, AI policy is expanding beyond industry and regulation into a matter of strategic national choice.
The policy difficulty of the AI era lies in the fact that technological, social and geopolitical uncertainties overlap and interact with one another. How far will the technology advance? How will that technology change society? How should states and the international community respond to those changes? Because technological development is difficult to predict, its social impact is also difficult to forecast. As a result, the responses of states and the international community become uncertain as well. At the same time, national policies and global competition structures influence the direction in which AI technology is developed and diffused.
Yet despite this uncertainty, policy judgment and response are still necessary. Even when the future cannot be accurately predicted, states must respond to present problems and set directions for the future. Those policy decisions will, in turn, influence the development of technology, markets and society. Therefore, the important task of AI policy is not to predict the future precisely and present one fixed solution. It is to design policy directions and social orders that can respond to change under conditions of uncertainty.
In this sense, the AI transformation should be understood not merely as technological innovation, but as a paradigm shift. Here, a paradigm means the basic framework of thought and order through which society, the economy, institutions and policy are organized. A paradigm shift therefore means a transformation in the way we understand the world, define problems, and use concepts, principles, institutions and policies to solve them. AI is changing existing assumptions about the relationship between humans and technology, between humans and labor, the meaning of creation and authorship, the production of information and knowledge, corporate and market competition, the sources of national competitiveness, and even the role of humans themselves.
If the internet transformed access to information, search engines transformed the discovery of information, and social media transformed the production and distribution of information, generative AI has begun to participate directly in the creation of information and knowledge. Agentic AI goes one step further by entering the realm of planning independently based on given goals, using necessary tools and executing tasks. The essence of the AI transformation, therefore, does not lie simply in the emergence of more powerful computers or new software. It lies in the fact that the very ways humans create information, accumulate knowledge, work, make things, exercise judgment and make social decisions are beginning to change.
This transformation also requires a new policy paradigm. Policy cannot simply wait because the direction of technological development is uncertain. Nor can it regulate the future rigidly based only on the current level of technology. It must promote innovation while responding to risk. It must strengthen corporate competitiveness while protecting the public interest. It must participate in the global AI ecosystem while securing strategic autonomy and AI sovereignty for the nation.
Ultimately, the central task of AI policy is not merely to respond individually to rapidly changing technologies. It is to determine what policy direction and principles should guide the design of a new social order in an age of uncertain AI transformation.
