The AI Basic Act, which includes measures to promote industry, has been passed in the National Assembly, and “there should be no further delays.” Over the next year, discussions regarding AI training data need to take place, as both the EU and the United States are currently engaged in similar discussions.
On the afternoon of the 26th, the first plenary session of the 420th session of the National Assembly was held in Yeouido, Seoul. During this session, the AI Basic Act was passed with 260 out of 264 members present voting in favor, 1 opposing, and 3 abstaining.
The Act is expected to take effect one year after being promulgated by the President. However, President Yoon Seok-yeol, who is required to promulgate the law, is currently suspended from duty, and Acting President Han Duck-soo, the Prime Minister, is also facing a potential impeachment motion. The political situation is becoming a variable for the implementation of the law.
Industry experts are concerned that any delay in the AI Basic Act could hinder the development of the industry. While the Act includes regulations on high-impact AI, its primary focus is on promoting technology development.
According to the Boston Consulting Group’s “AI Maturity Matrix” report, South Korea’s AI technological maturity and potential are ranked 6th globally, classified in the second tier.
Rep. Lee Hoon-ki of the Democratic Party of Korea stated, “We cannot afford to delay in the global AI race,” adding that, “Even amidst martial law and impeachment circumstances, the National Assembly should strive to pass critical legislative tasks for the welfare and development of the industry.”
During the grace period, various ministries will need to prepare detailed enforcement ordinances and guidelines for the implementation of the law. Discussions about the disclosure of AI training data are expected to intensify.
The Ministry of Culture, Sports, and Tourism has requested the addition of a clause for the disclosure of training data lists required for developing generative AI. They argue that clear regulations on copyright and ownership of AI-generated content are necessary.
Five media organizations, including the Korea Newspaper Association, are urging the National Assembly and the government to legislate the mandatory disclosure of training data by AI businesses. They argue that failing to disclose this data severely infringes on the rights of copyright holders.
The European Union’s AI Act already stipulates related matters about the copyright of training data. It requires the disclosure of details about copyrighted works used to train general AI models released to the market. This allows copyright holders to exercise their rights in the future.
In the United States, a “Bill on the Disclosure of Training Data” was introduced last April. It requires that works used in building generative AI systems be reported to the Copyright Office, allowing copyright holders to clearly know if their works have been used for AI training and manage their use against unlawful exploitation.