"Bought Hundreds of Thousands of Korean Books for AI Training?" ... Choi Hwiyeong: "Current Law Limits Response"

Anthropic’s Bulk Purchase of Korean Books Draws Response at National Assembly
21 Works by Korean Authors, Including Han Kang and Hwang Sok-yong, Included in U.S. Copyright Settlement
Buying Books Does Not Grant Permission for AI Training... Need for Legal Reforms Acknowledged

Choi Hwiyeong, Minister of Culture, Sports and Tourism, acknowledged the need to improve the legal framework, saying current laws offer limited ways to respond to the controversy over U.S. artificial intelligence (AI) company Anthropic’s bulk purchases of Korean books and its unauthorized use of them for AI training. After evidence emerged that works by Korean authors had been used to train AI without the copyright holders’ consent, reports that hundreds of thousands of Korean books had been exported to the United States further brought copyright protection for publications used in AI training into focus at the National Assembly.

Minister of Culture, Sports and Tourism Choi Hwiyeong delivers the government's proposal at a plenary meeting of the National Assembly's Culture, Sports and Tourism Committee on the 5th. Yonhap News

At a plenary meeting of the National Assembly’s Culture, Sports and Tourism Committee on the 5th, Minister Choi responded to a question from Rep. Choi Min-hee of the Democratic Party of Korea about Anthropic’s bulk purchase of Korean books and their use for AI training, saying, “Under current law, there is no way to respond. Improvements are needed.”

He said measures must be developed to address AI training conducted without copyright holders’ consent, adding, “This is an issue that needs to be discussed with the World Intellectual Property Organization (WIPO).”

Anthropic, at the center of the controversy, is a U.S. company that developed the generative AI chatbot Claude. The publishing industry has increasingly protested since reports emerged that Anthropic had purchased large quantities of Korean-language books through domestic book distributor Woongjin Book Center.

An analysis of import and export shipping records also found that about 166 metric tons of books had been shipped to the United States through Woongjin Book Center from last year through June this year. Assuming an average weight of 400 grams per book, that amounts to about 410,000 books. U.S. bookseller Better World Books was involved in sourcing the books, and Woongjin Book Center confirmed that it had signed a book supply contract with Anthropic. The complete list of books purchased and whether individual titles were used for training have not been disclosed.

Anthropic’s copyright issues first came to light through a class-action lawsuit brought by authors in the United States. In June 2025, a U.S. court ruled that Anthropic’s scanning of print books it had legally purchased and use of them for AI training constituted fair use. However, the court did not consider it fair use to obtain and retain books downloaded from pirated websites.

Anthropic later reached a settlement with the authors for a total of $1.5 billion (about 2 trillion won), which received final approval from a U.S. court in July this year. The settlement covers about 500,000 works, with compensation of about $3,000 per work.

Works by Korean authors were also included in the settlement. Last month, the Literature Translation Institute of Korea identified 21 works by Korean authors on the Anthropic copyright settlement website and informed the authors about the matter and available remedies.

The list included Human Acts by Nobel Prize-winning author Han Kang, as well as Hwang Sok-yong’s The Guest, Kim Young-ha’s The Empire of Light, Shin Kyung-sook’s Please Look After Mom, Kim Ae-ran’s My Brilliant Life, and Hwang Sun-mi’s The Hen Who Dreamed She Could Fly. These works were distributed overseas, including in English editions, and their inclusion stemmed from a separate case involving the acquisition of pirated copies, unrelated to the recent bulk purchase of print books in Korea.

There have also been continued efforts to purchase large quantities of books domestically, raising suspicions that they may be intended for AI training. Since early this year, Kyobo Book Centre has received requests from 10 overseas companies, including companies from the United States and the United Kingdom, each seeking to purchase one copy of about 100,000 different titles. Kyobo Book Centre declined to supply the books, citing the unclear purpose of the purchases and final delivery destinations, as well as the possibility that the books would be used for AI training. In August, YES24 also received an inquiry from a U.S. bookseller about purchasing used books in bulk.

The publishing industry maintains that buying a book legally does not automatically grant the right to reproduce and use the work for AI training. South Korea’s Copyright Act allows the resale of books that have been legally sold, but reproducing them as digital files or using them for AI training may require a separate copyright assessment. Under current law, questions about fair use, the scope of usage permissions, and how to enforce rights against overseas businesses remain unresolved.

The Korean Publishers Association has begun surveying bookstores, publishers, and book distributors about purchases of books suspected of being intended for AI training. The Korean Publishers Society has also issued an appeal titled “No Books, No AI,” calling on AI companies to disclose how they use training data and to provide fair compensation to copyright holders.

The publishing industry’s central demand is for institutional mechanisms that make it possible to identify the sources of books used for AI training and how they were used. Without knowing which AI models have used their books, authors and publishers say, it is difficult to decide whether to grant permission or seek compensation for copyright infringement.

The government has also been discussing institutional improvements concerning the use of AI training data and copyright protection. The Ministry of Culture, Sports and Tourism is operating a consultative body involving copyright holders, AI developers, and legal experts to consider disclosing training data, securing lawful usage rights, and facilitating transactions involving copyrighted works.

Minister Choi acknowledged the limitations of current law today and said measures must be developed, including through discussions with international organizations.

This content was produced with the assistance of AI translation services.