Home > National > Social Affairs

Gov't committee faces backlash after 'use first, settle later' AI data training proposal

Published: 29 Dec. 2025, 19:08

Im Moon-young, vice chairperson of the National AI Strategy Committee, fourth from left, and attendees pose for a photo during a press briefing marking the committee’s 100th day since its launch at Seoul Square in Jung District, central Seoul, on Dec. 15. [YONHAP]

Backlash is growing over a proposal by a government committee to allow AI developers to use data for training without obtaining prior permission from creators, under a so-called “use first, compensate later” system. Critics warn the move could distort fair compensation structures and further weaken the creative industry.

The National AI Strategy Committee is discussing amendments to the Copyright Act with relevant ministries, including the Ministry of Culture, Sports and Tourism and the Ministry of Science and ICT, according to AI and content industry sources on Sunday. In a draft of the “Korea AI Action Plan” released on Dec. 15, the committee recommended that the Culture Ministry prepare a revision to the Copyright Act by the second quarter of next year.

“Due to legal uncertainty, AI developers are currently deciding not to use data at all,” said Baek Eun-ok, a professor in the Department of Computer Software Engineering at Hanyang University and head of the committee’s data subcommittee, adding that the idea is to move away from a “permission first, use later” model to a “use first, settle later” approach.

“This means broadly allowing AI training while building a system that ensures fair compensation for creators,” she said.

Until now, the “use first, settle later” approach under copyright law has been allowed only on a limited basis for special purposes, such as educational materials. The committee is now considering applying it to AI training in cases where copyright holders are unclear.

This would include data publicly available on the web whose rights holders are difficult to identify, as well as works whose copyright holders are known but have not explicitly opted out of data collection.

Although the proposal is technically a recommendation, it is likely to be pursued as is once the action plan is finalized at a plenary session chaired by President Lee Jae Myung himself.

Park Hak-ki, vice chairman of the Korea Music Copyright Association, center, reads a joint statement during a press conference calling for the protection of creators’ rights in the AI era at the Korea Broadcasters Association Hall in Mok-dong, Yangcheon District, western Seoul, on May 14. [YONHAP]

“Waiting for copyright negotiations to naturally take place in the market could take several more years,” said a committee official. “Overseas AI models are already known to have trained on virtually all publicly available internet data. There was a strong sense of urgency that we had to try everything we could.”

Even if the measure is seen as necessary to keep up in the global AI race, backlash from the content industry is expected to intensify. Creative groups in the literary, fine arts, broadcasting, photography, film and music sectors have already called for a fair compensation system, saying creators often do not even know how their content is being used for AI training.

“The government and industry are unilaterally forcing creators to make sacrifices,” said a representative from the Korea Music Copyright Association. “If payments are framed as compensation rather than usage fees, creators’ bargaining power will weaken, raising concerns over whether fair compensation will actually be achieved.”

“If content can be used without permission, who would want to produce high-quality works?” said Kim Si-yeol, executive director of the Korean Publishers Association, adding that disputes between creators and companies could become even more costly.

Legal experts have also raised questions about whether AI training truly serves public interests. “There needs to be social consensus on whether AI development has public value comparable to education,” said Choi Sung-jai, a professor of law at Sejong University, warning that overseas companies could end up profiting from training on Korean content.

The OpenAI logo is displayed on a mobile phone in front of a computer screen with output from ChatGPT in Boston on March 21, 2023. [AP/YONHAP]

Choi cited OpenAI as an example, noting that ChatGPT generated significant revenue by producing AI images in the style of Japan’s Studio Ghibli, without any compensation to the rights holders. Japan is one of the countries that has adopted a “text and data mining” (TDM) exemption, allowing the use of copyrighted works without permission or compensation as long as it does not unreasonably harm the copyright holder’s interests.

News content is also being used for AI training without fair compensation. In April, the Korea Newspaper Association filed a complaint with the Fair Trade Commission against Naver, accusing it of unfairly using news content for generative AI training and abusing its market dominance. According to data disclosed during an October parliamentary audit by conservative People Power Party Rep. Choi Su-jin, news content accounted for 13.1 percent of the data Naver used for AI development.

The committee plans to collect public feedback on the AI action plan draft through Monday. To address concerns from the content industry, alternatives such as opt-out mechanisms, enhanced transparency in AI training and safeguards to prevent reproducing original works have been proposed. The idea is to block AI training on content whose copyright holders have opted out and to disclose training scopes to build an AI ecosystem that benefits creators.

The debate over copyright exemptions for AI training is also intensifying overseas. While the European Union allows TDM exceptions even for commercial entities, a German court ruled last month that OpenAI violated copyright law by training its AI model on song lyrics. The British government also scrapped plans in 2022 to allow TDM for commercial purposes following strong opposition from the cultural sector.

In contrast, U.S. courts have issued rulings that using books and news content without permission for AI training can qualify as fair use. However, controversy continues as prominent authors have filed copyright infringement lawsuits against OpenAI and other AI developers.

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
BY CHANG YOON-SEO [[email protected]]