[AI Rulemaker]②Lee Haemin: "I Created a Law to Prevent a Dystopian AI Era"

Standing at the Crossroads of Dystopia and Utopia

Strengthening Software Is Essential for Becoming a Top Three AI Power

AX and Security Now Mean the Same Thing

Editor's NoteFollowing the European Union (EU), South Korea became the second country in the world to enact a comprehensive Artificial Intelligence (AI) Basic Act. We meet the politicians who have spearheaded AI-related legislation, explore their philosophies on shaping an AI-driven nation, examine legislative follow-up tasks, and seek strategic and regulatory solutions to survive in the global AI race.

"Noah's Ark." In the Bible, this was the only vessel granted to humanity when God decided to flood the world and create a new one.

Hae-Min Lee, a lawmaker of the Party for National Innovation, repeatedly brought up "Noah's Ark" during the hour-long interview with The Asia Business Daily. It revealed his concerns about a world undergoing a shift to the AX (AI Transformation) era.

Assemblyman Lee remarked, "It is said that after the Industrial Revolution, it took 200 years for human productivity improvements to become fully established, but we are now experiencing such changes in just a few decades. If left unchecked, the inevitable result will be a K-shaped growth marked by extreme wealth polarization. We need a system that can steer us toward utopia, rather than slipping into dystopia." He went on to introduce this law as 'a bill designed to change the course of society at the crossroads between dystopia and utopia, in an effort to avert a dystopian future.'

Heomin Lee, a member of the National Assembly from the Party for National Innovation, is being interviewed by The Asia Business Daily in his National Assembly office. August 12, 2026. Photo by Hyunmin Kim

The AI Transformation Response Act obligates the government to establish a national strategy for AI-based industrial transformation as well as a basic plan in five-year intervals. It also explicitly stipulates the rights of citizens to receive job retraining, transition support, and minimum livelihood security. Regions experiencing particularly severe shocks from industrial transformation are to be designated as "transition areas," receiving tailored support measures. The Act also requires companies to develop worker retraining plans and to provide advance notice and consultation in cases of large-scale employment changes. Companies that demonstrate good-faith compliance may receive tax benefits.

The two main operational mechanisms of this law are "employment" and "funds." Assemblyman Lee explained, "The more a company maintains employment, the more advantageous it is for the company. When layoffs occur due to AI adoption, companies must contribute to a fund accordingly, while firms that create more jobs are rewarded." He also noted there are separate provisions to protect startups and small-to-medium enterprises. Assemblyman Lee emphasized that the philosophy behind the bill is not about regulating businesses, but about ensuring the ongoing existence of industry.

Assemblyman Lee agrees with the current administration's focus on becoming one of the "world's top three AI powers" as its core national agenda, but he also questioned whether the direction is clearly defined.

He said, "I acknowledge that the current administration has set AI as the most significant policy priority, but we have to recognize the challenges that come along with it." Assemblyman Lee went on, "I asked the government for a clear definition of 'the three leading AI nations,' but received no concrete answer. Using AI chatbots, I inferred that the value chain includes not only datasets used to train models, GPUs, and HBM, but also data centers (including cooling and power supply)." He added, "I believe we are trying to maximize our ecosystem using our strengths in memory semiconductors, especially HBM."

Assemblyman Lee remarked, "Of course, it is impossible to excel in every stage of the value chain, but the government is making efforts to secure a strong presence in areas such as physical AI and AI data centers." However, he also projected that time is running short. While much of the current focus is on hardware, he believes the ultimate goal should be advancement into the software sector.

He said, "The first profits always start from infrastructure—hardware. That's how companies like Nvidia, TSMC, SK hynix, Micron, and Samsung Electronics are making money," but added, "Next, leadership will inevitably shift to high value-added software." He went on to point out that as China rapidly catches up in hardware, the main challenge is whether Korea can also make headway into software.

Heemin Lee, a member of the Party for Democratic Innovation, is being interviewed by The Asia Business Daily in his National Assembly office. 2026.8.12 Photo by Hyunmin Kim

The following is a Q&A.

-There are concerns that the AI Transformation Response Act will increase burdens on companies.

▲If consumers disappear, companies disappear too. In the short term, productivity grows rapidly and employers might believe it's good to cut people. Machines can work 24/7 at consistent quality, and if their capabilities improve, you could replicate them endlessly and have a million machines working identically. Humans are not like that. If we only pursue productivity gains, we lose social sustainability, and eventually, industries themselves cannot survive. The law was designed with this philosophy in mind. Our 200 years of experience since industrialization show us that, if left alone, the K-shaped gap will widen. Knowing this, we must prepare in advance—fixing it later would entail tremendous social costs.

-In the latter half of the National Assembly term, you moved from the Science, ICT, Broadcasting, and Communications Committee to the Political Affairs Committee.

▲Now that we've built the "framework" (the AI Basic Act), it's time to move toward AX, and I thought everything could be managed within the Science, ICT, Broadcasting, and Communications Committee, but I found out that wasn't the case. Issues like personal data protection and AI transformation in the finance sector are all related to AI, but the relevant institutions are scattered. Both finance and defense are important, and since the Personal Information Protection Commission belongs to the Political Affairs Committee, I decided to start there. At times, it feels like I'm serving on two committees (Political Affairs and Science, ICT, Broadcasting, and Communications), so I'm grateful and sorry to my office staff.

-What issues will you prioritize in the Political Affairs Committee?

▲Personal information protection, AX in the finance sector, and AI agents. The core principle beneath these is the same: 'AX and security now mean the same thing.' Financial firms are particularly proactive about AX, not merely for efficiency, but primarily due to security. Given that AI models searching out and exploiting each other's weaknesses has become everyday reality, the only way to defend against the most powerful hacking threats is with equally strong AI.

-Should the National Assembly's organizational structure for AI-related legislation change?

▲For AI-related matters, there needs to be an organization like the Judiciary Committee that comprehensively oversees everything. If we can't create another standing committee or a special committee, all bills are routed to the Judiciary Committee, so at the very least, a dedicated subcommittee should be created within it. Things need to be streamlined. If we continue as we are, I worry the Judiciary Committee won't be able to resolve conflicts if contradictory bills are submitted. Even if everything looks fine legally, there may be industrial ramifications like, "What is this supposed to be?" If not, then each standing committee should at least have—or dispatch—two AI, IT, or digital experts as members.

-How do you evaluate the AI Basic Act, which passed last year?

▲It was proposed in December 2024, passed standing committee and plenary sessions immediately, and came into effect in January this year. It's significant that Korea was the second in the world to legislate such an act, and the first to implement it. Unless you try these things, you never find out what doesn't work and what suits reality. Now, as we operate the implementing regulations, we are accumulating a year's worth of evidence about what works well and what does not; this trove of knowledge will become invaluable data that can't be found anywhere else in the world. The AI Basic Act is, as the name suggests, a framework law focused on defining governance and setting up the foundation—what to fill it with is being fleshed out through the implementing regulations.

-What was the most regrettable part of the legislative process?

▲The subcommittee review was very tough. Nineteen different bills were merged for review, and in the end, 85-90% of the final version was from my bill, but the remaining parts I couldn't stop caused issues. The problem was with the definition of "high-impact AI." The original draft defined it as "AI that has a significant influence on the safety and fundamental rights of life or the body"—an "and" clause. Other committee members with backgrounds in Korean literature or journalism argued that it should be "AI that has a significant influence or poses a risk," turning it into an "or" clause and adding the notion of "risk." This dramatically broadened the scope, and the law was criticized in the press for being overly restrictive. Originally, my intention was to sharply regulate only those areas directly impacted, and relax everything else, but the final law ended up doing the opposite.

-Why was the 'and' changed to 'or'?

▲The reasoning was that it was grammatically better. There were members with backgrounds in Korean literature or the media, and as an engineering graduate, grammar was not my forte. They said, "Long 'and' clauses are not desirable." Thankfully, at least they understood the legislative intent. I kept repeating, "legislative intent, legislative intent" in the transcripts during the subcommittee meetings. Thanks to this, the Ministry of Science and ICT had a very hard time narrowing the expanded scope when drafting the regulations.

-Why coin a new concept 'high-impact'?

▲The EU uses "high-risk" because it is focused on regulation. Canada, on the other hand, uses "impact," reflecting a more promotional stance, as do the United States and Canada. I believed the law needed to be viewed fully from the user’s perspective and thus chose "impact." Previous legislation considered only three stakeholders: model developers, service providers, and paying users. I introduced the concept of "those influenced by AI." Using autonomous vehicles as an example: engine manufacturers, finished vehicle/service providers, users such as Tesla owners—and then there are bystanders affected by the vehicles. These people did not participate in any earlier phase, but if there is an accident, they are affected all the same. In the case of AI-powered medical devices, patients diagnosed by doctors using the device should also be considered. When tracing back from an incident, it became clear that responsibility could lie with any number of parties.

This content was produced with the assistance of AI translation services.