[Editorial] As the 'AI Basic Act' Becomes the First in the World to Take Effect, the Government Must Prevent Confusion for Businesses

On the 22nd, the Framework Act on the Development of Artificial Intelligence and Establishment of Trust (AI Basic Act) will come into full force. This will make our country the first in the world to put a comprehensive AI regulatory law into full operation. We stand at an inflection point where AI is transforming paradigms across the economy and society as a whole. At a time of heightened industrial instability, putting an institutional framework in place is a welcome step. By creating a foundation for AI development through institutional support, we can also enhance national competitiveness. However, we must not become so fixated on the label of being the first in the world that we ignore concerns on the ground. The European Union (EU) was in fact the first to adopt an AI law, but it has postponed its implementation. The reason the EU opted for phased implementation is likely that it carefully weighed the burdens the AI law would impose on industry.
Even in our country, serious conflict has arisen between the government and industry ahead of the AI Basic Act’s enforcement. The government stresses that the AI Basic Act is not a regulation-centered law but a tool to promote industry. Even so, in light of industry discontent, it has announced that it will minimize the exercise of its fact-finding authority in cases of legal violations and allow a guidance period of at least one year before imposing administrative fines. Furthermore, the government highlights that once the AI Basic Act takes effect, the previously uncertain AI regulatory environment will actually become more transparent. The argument is that by establishing a clear system and guidelines, the law will help companies make business and investment decisions.
However, there is a wide gap between the government’s assessment and the reaction on the ground. According to a survey by Startup Alliance, 98% of domestic AI startups responded that they have not established a practical response system for when the AI Basic Act is fully enforced. As for the government’s plan to grant a grace period before imposing fines, the prevailing sentiment is that this shows a lack of understanding of the reality on the ground. Companies lament that once they are branded as lawbreakers, they lose trust in this field and their business itself becomes difficult to sustain.
An even bigger problem is the possibility that only domestic companies will end up being disadvantaged. Even if global AI big tech firms distribute deepfakes or harmful content in the Korean market, there are clear limits to applying domestic law to them. While they are required to designate a local representative, in reality it is not easy to hold them substantively accountable. Compared with domestic companies, which become direct targets of regulation and punishment, foreign firms are effectively left in a legal blind spot. In the end, there is a risk that an unfair structure will become entrenched in which, amid global competition, only domestic companies are hobbled by excessive regulation.
Even at this stage, the government must listen carefully to the voices of industry. If the law goes into effect as is, there are concerns that the confusion seen before the enforcement of the Serious Accidents Punishment Act will be repeated. At that time as well, the government promised sufficient preparation time and guidance. Yet once the law took effect, companies had to pour enormous resources into legal consultations and building safety management systems. The AI Basic Act likewise contains many provisions whose interpretation is disputed, such as the vague criteria for "high-impact AI" and watermarking requirements. Similar side effects to those seen with the Serious Accidents Punishment Act could arise under the AI Basic Act. Therefore, the government should consider, even after implementation, measures to accommodate ventures and startups that have weak capacity to respond to the law and limited capital.
In addition, once the law takes effect this month, flexible enforcement by the government will be all the more necessary. Given the many ambiguities in the law, it should not be applied rigidly and uniformly according to a fixed yardstick; instead, it must be implemented flexibly in light of the difficulties faced on the ground. Going forward, we hope the government will fully incorporate industry feedback during actual enforcement and use it to remedy the shortcomings of the AI Basic Act.