School’s in for AI: Academic papers fair game for training under guideline, news likely not
Published: 26 Feb. 2026, 20:59
Updated: 27 Feb. 2026, 12:55
-
- KIM MIN-YOUNG
- [email protected]
The ChatGPT logo is seen in this photo taken on Jan. 22, 2025. [REUTERS/YONHAP]
The government has determined that developing AI models using open-access academic papers may qualify as fair use, but doing the same with news articles and offering summary services without permission from media outlets is unlikely to be recognized as such.
The government on Thursday released new guidance outlining how the Copyright Act’s fair-use provision may apply when copyrighted works are used to train generative artificial intelligence models.
The document is intended as a reference for copyright holders, as well as AI developers and users.
The move follows a policy discussion at the National Museum of Modern and Contemporary Art in central Seoul on Thursday morning, which was attended by Culture Minister Chae Hwi-young, Deputy Prime Minister and Science and ICT Minister Bae Kyung-hoon and National AI Strategy Committee Vice Chair Im Moon-young, according to the government’s release.
The guide explains the four factors typically considered in fair-use reviews: why an individual is using a copyrighted work; the nature of the copyrighted work; the amount of the copyrighted work that the person is using; and the effect that the individual's work will have on the copyrighted work's current or potential market.
The guide also presents hypothetical examples to illustrate which circumstances count as fair use. However, it emphasizes that any conclusions depend on the facts of each case and will ultimately be made in court.
The OpenAI logo is displayed on a mobile phone in front of a computer screen with output from ChatGPT on March 21, 2023 in Boston. [AP/YONHAP]
In addition to providing the guide, the government aims to strengthen the infrastructure for lawful content use in AI training.
The Ministry of Culture, Sports and Tourism plans to expand systems that help developers identify rights-management information and enter licensing agreements to reduce legal transaction costs.
The Ministry of Science and ICT will link rights-information systems with private data marketplaces through an integrated AI training data framework and consider measures to encourage licensed data transactions, including tax incentives tied to requesting and allowing copyrighted works for AI training.
A computer screen in English teacher Casey Cuny's classroom shows ChatGPT during class at Valencia High School in Santa Clarita, California on Aug. 27, 2025. [AP/YONHAP]
The government also pointed to recent changes to the Korea Open Government License (KOGL), which governs the use of public-sector works. KOGL introduced a Type 0 label allowing unrestricted use and an AI Type label that permits use for AI training under specified conditions.
To address disputes, the Korea Copyright Commission is operating an AI-focused consultation channel that provides legal guidance — including via its hotline at 1800-5455, reached by pressing 0 — on copyright issues related to AI training.
“We will continue to update the guide to reflect new case law and technological developments,” said Culture Minister Chae. “We will work to keep a balance between protecting creators and enabling lawful use of works in AI models.”
BY KIM MIN-YOUNG [[email protected]]





with the Korea JoongAng Daily
To write comments, please log in to one of the accounts.
Standards Board Policy (0/250자)