South Korean publishers are raising concerns over large-volume book purchase requests that they suspect are linked to artificial intelligence training, prompting calls for enhanced copyright protections and fair compensation.

Since early 2026, several overseas companies, including firms from the United States and the United Kingdom, have reportedly approached bookstores such as Kyobo Book Centre with requests to buy single copies of approximately 100,000 different titles. A Kyobo spokesperson indicated that many of these buyers identified themselves as book wholesalers, while some denied intentions related to AI training but acknowledged that data-related companies were the end recipients. Kyobo declined these bulk orders, citing uncertainty about potential use for AI model development.

Yes24, another major South Korean bookstore, also received a similar inquiry in late August from a U.S. distributor seeking large volumes of used books. Initial discussions ceased after the distributor stopped responding.

The issue has drawn heightened attention following revelations by the Literature Translation Institute of South Korea. The institute recently informed South Korean authors that 21 of their works appeared on a compensation eligibility list arising from a U.S. copyright settlement involving Anthropic, an AI company facing a class-action lawsuit accusing it of illegally downloading copyrighted books for training purposes.

Concerns over bulk purchases suspected of AI-related scanning are not unique to South Korea. In Japan, there have been reports of large-scale acquisitions of second-hand books in fields such as philosophy, history, and medicine, intensifying suspicions that AI developers seek extensive textual material for model training.

In response, the Korean Publishers Association has initiated a survey targeting bookstores, distributors, second-hand booksellers, and publishers to collect data on bulk orders potentially connected to AI. An association representative indicated plans to establish a collaborative body aimed at advocating for compensation frameworks and relevant legislation once the survey is complete.

The Korean Publishers Society has also issued a public appeal, signed by over 240 publishers and organizations along with more than 7,000 authors, translators, illustrators, and other publishing professionals, demanding equitable remuneration when works are used for AI training.

Industry figures emphasize challenges in addressing these concerns. Chon Hana, head of publisher Prunsoop, highlighted that individual rights holders struggle to negotiate with large foreign technology companies. Publishers also face difficulties obtaining transparency on how and where their content is acquired and utilized by AI developers.

Park Jung-in, a copyright expert and professor at Duksung Women’s University’s AI DynaInfo Institute, pointed to the information gap as a significant hurdle, complicating licensing negotiations and infringement claims. Park called for establishing mechanisms that would enable rights holders to verify sources and applications of training data, describing the problem as a systemic issue where rights exist but are hard to enforce in practice.