[AI Safety Crisis] "Korea, If Delayed Further, Global Entry Will Be Difficult"... EU and US Aiming to Become 'Rule Makers'

"AI Framework Law Must Be Hastened for Global Competition"
EU and US Set Rules with AI Laws and Executive Orders
South Korea's AI Framework Law Pending in National Assembly

"If norms for artificial intelligence (AI) are not established, domestic companies may find it difficult to compete in the global market."

There is a growing call to put safety measures in place for rapidly advancing AI technology. As institutional frameworks are being established, led by Europe and the United States, there is an urgent need to prepare legislation domestically as well. Falling behind global standards could limit the development and growth of the domestic AI industry. Kim Myung-joo, President of the International Artificial Intelligence Ethics Association and Director of the Barun AI Research Center at Seoul Women’s University, emphasized, "Domestic AI companies, which have been in a regulatory blind spot, may encounter unexpected obstacles when entering overseas markets such as Europe. It is necessary to establish common norms to avoid having to redevelop AI technology to comply with foreign regulations or to be blocked."

[AI Safety Crisis] "Korea, If Delayed Further, Global Entry Will Be Difficult"... EU and US Aiming to Become 'Rule Makers'

Kim Myung-joo, President of the International Artificial Intelligence Ethics Association (Director of the Barun AI Research Center at Seoul Women's University)

Europe, ahead of us, is preparing the so-called 'AI Act,' which will be fully enforced starting in 2026. It classifies AI-related risks into four categories: minimal risk, limited risk, high risk, and unacceptable risk, with regulations corresponding to each. In particular, limited risk covers chatbots and deepfake risks, requiring disclosure that the content is AI-generated. Infrastructure-related AI is classified as high risk, imposing transparency obligations through security, management, and tracking systems. Unacceptable risk targets AI related to subconscious manipulation of humans and biometric classification systems, which are generally prohibited except for serious crimes.

The United States also established an 'Executive Order on Safe and Trustworthy AI' in October last year. U.S. companies must notify the federal government if their developed AI poses risks to national security, the economy, or health, and share safety test results and key information about the AI with the government. Federal agencies are required to disclose content when using AI and verify and monitor side effects.

President Kim said, "At first glance, the EU’s AI Act seems solely regulatory, but it also includes provisions for promotion. Because AI technology is advancing rapidly, it seems desirable to first define it abstractly, as in Europe’s AI Act, and then develop detailed rules through enforcement regulations."

Europe, the United States, and others have begun full-scale AI-related regulations. In particular, the UK and the U.S. have signed a bilateral agreement to cooperate in checking AI safety knowledge and information. However, domestically, not only is legislation lacking, but even parliamentary discussions are not progressing properly. The AI Basic Act (Act on the Development of Artificial Intelligence and the Establishment of a Trust-Based System) has been proposed but remains pending in the National Assembly. If it does not pass the plenary session before the end of the 21st National Assembly’s term, it will be automatically discarded.

AI startups were cautious about preemptive regulation but agreed on the need for legislation to establish a foundation. An AI startup CEO, who requested anonymity, said, "We are at a stage where we need to prove the economic value of AI for the investments made so far, and regulations could undermine the market and company growth before they are established." However, he added, "A policy direction of allowing first and regulating later, along with private sector self-inspection, is necessary, but from another perspective, legal and administrative foundations for AI commercialization are indeed needed, so legislation reflecting this is required."

While the enactment of the domestic AI Basic Act is delayed, the upcoming 'AI Seoul Summit' jointly hosted by the Korean and UK governments on May 21-22 could be a turning point. It is a follow-up meeting to the 'AI Safety Summit' held in the UK last November, where AI safety, innovation promotion, and concrete cooperation plans will be discussed and shared. President Kim said, "Since the UK has left the European Union (EU) due to Brexit, it is currently in a similar position to us, caught between the EU and the U.S. in terms of AI-related legislation. Because both countries are in similar situations, they can align their efforts, and constructive meetings on regulatory directions and industrial promotion are possible."