GPT-3, introduced in 2020, was an important technological turning point that demonstrated the full potential of large language models. Yet the decisive moment when AI began to move beyond the domain of researchers and developers and into the everyday lives of the general public came on November 30, 2022, when ChatGPT, based on the GPT-3.5 family of models, was released to the public. As anyone became able to converse with AI in natural language, receive answers to questions, and generate writing and software code, artificial intelligence rapidly began to shift from a specialized technology into a general-purpose intellectual tool.
In just a few years, generative AI has moved beyond laboratories and advanced technology companies and has quickly entered the daily lives of ordinary citizens, corporate activity, education, labor, media, culture, public administration and national strategy. Like the internet and smartphones before it, AI first appeared as an innovative technology. But it is now evolving into a general-purpose technology that changes the way society itself operates.
The changes that have unfolded in just three years since the emergence of generative AI show the speed of this transition in compressed form. According to Stanford University’s AI Index Report 2026, generative AI spread within roughly three years to the point where about 53% of the U.S. population aged 18 to 64 used it. Corporate adoption of AI has expanded even more rapidly. The share of companies using AI rose from 55% in 2023 to 78% in 2024 and 88% in 2025. AI is no longer an experimental technology used only by a handful of technology companies or innovative early adopters. It is becoming a general-purpose tool rapidly embedded in individual daily life and corporate management. Individuals use AI to search for information, write, translate, create images and videos, and develop software. Companies are also rapidly expanding AI use across customer support, marketing, software development, research and development, and even managerial decision-making.
What matters is not simply the increase in adoption rates. AI is beginning to change the structure of real economic activity and the capabilities required in the labor market. According to PwC’s 2025 Global AI Jobs Barometer, revenue per employee in industries highly exposed to AI increased by 27%, about three times higher than the 9% growth seen in industries with low AI exposure. Jobs requiring AI skills carried an average wage premium of 56%, and in occupations highly exposed to AI, the skills demanded by employers were changing 66% faster than in other occupations. This shows that AI is moving beyond a tool for improving the efficiency of existing work and is reshaping the demand structure of the labor market itself, including what kinds of talent and capabilities companies need.
At the level of individual tasks, AI’s productivity effects are also becoming visible. In customer support, the use of AI has been associated with productivity gains of around 14 to 15%. In software development, studies have reported gains of about 26%, while marketing content production has shown productivity or output improvements of around 50%. These figures must be interpreted carefully because they vary depending on industry characteristics, task types and measurement methods. Nevertheless, it is clear that AI has already begun to intervene in actual work processes and change the division of roles between humans and machines.
Early signs of change are also appearing in employment. The World Economic Forum projects that about 22% of all jobs will be affected by structural change between 2025 and 2030. During the same period, it expects 170 million new jobs to be created and 92 million jobs to be displaced, resulting in a net increase of about 78 million jobs. Therefore, it is not appropriate to explain labor-market change caused by AI simply as “job destruction.” The more important change lies in job transformation, in which the content of existing roles and the capabilities required for them are rapidly reconfigured. The WEF also projects that about 39 to 40% of the core skills required in current jobs will change by 2030.
These changes show that AI is no longer merely software that assists workers. It is becoming a technology that determines anew which tasks humans should perform, which tasks AI should perform, and what new capabilities humans will need. AI’s impact is not confined to the AI industry itself. It is rapidly spreading across companies, industries, labor markets, education, national competitiveness and social institutions.
This is precisely where AI becomes a matter of policy. In general, a new technology becomes a policy issue when its effects move beyond the private relationship between developers and users and begin to influence society as a whole. The rise of automobiles created traffic laws and road policy. The emergence of broadcasting created spectrum policy and broadcasting regulation. The spread of the internet led to the development of telecommunications policy, platform policy, privacy protection and cybersecurity policy.
AI is following the same path, but its scope is far broader. AI affects not only labor markets and productivity, but also the production and circulation of information, education, healthcare, finance, defense, public administration, scientific research, cultural creation, elections and public opinion, as well as human judgment and decision-making. As a result, the development of AI can no longer be explained or managed solely through the autonomous choices of technology developers and markets. Both the benefits and risks created by AI have begun to spread throughout society.
The question, therefore, must also change. It is shifting from “How far should AI technology be advanced?” to “How should we design a society in which AI operates?” This is the most fundamental reason AI policy has emerged.
