Watermarking AI-Generated Content Mandatory in US and EU... What About Korea?

There is a growing call for the introduction of regulations, such as mandatory labeling of AI-generated content, to reduce deepfake damage. While Europe and the United States have already implemented regulations, related discussions in South Korea remain at a standstill.

According to industry and political circles on the 7th, currently, if deepfakes are used to create pornographic material that causes sexual humiliation, it can be punished under the Sexual Violence Punishment Act. However, producing pornographic material using deepfakes itself is not subject to punishment. It is also difficult to prevent deepfakes from being misused for defamation or financial fraud beyond pornography.

원본보기 아이콘

The government and political circles believe that regulatory measures should be established by mandating labels such as watermarks on AI-generated content. The idea is to strengthen accountability through mandatory labeling, considering both the promotion of the AI industry and the protection of users.

However, the handling of the AI Basic Act, including the mandatory watermarking, is sluggish. After passing the bill subcommittee of the National Assembly’s Science, Technology, Information and Broadcasting Committee in February last year, it has been pending in the full committee meeting for nearly a year. This is because civic groups mainly oppose the core of the AI Basic Act, ‘prior permission and post regulation.’ The ‘prior permission and post regulation’ means that AI technology development should not be restricted unless it harms life, safety, or rights and interests.

Meanwhile, amendment bills have also been proposed. Lee Sang-heon, a member of the Democratic Party of Korea, proposed an amendment to the Content Industry Promotion Act in May last year, mandating labeling of AI-generated outputs. In December last year, Kim Seung-soo of the People Power Party proposed a similar amendment to the Information and Communications Network Act. The amendments still have contentious issues regarding the scope and methods of labeling and enforcement measures. Some argue that it could even discourage cases where AI is used simply as a tool in the creative process or without malicious intent.

Overseas, moves have been faster. The European Union (EU) mandated separate labeling for AI-generated content through the Digital Services Act (DSA) in August last year. The U.S. administration also strengthened AI content identification measures through an executive order in October last year.

Professor Kim Myung-joo of the Department of Information Security at Seoul Women’s University said, "Except for text-generated content, technically, it is not difficult to insert watermarks," adding, "Rather than enacting the AI Basic Act, which requires social discussion, mandatory watermarking through amendment bills should be implemented to reduce deepfake damage."

© The Asia Business Daily(www.asiae.co.kr). All rights reserved.