Anthropic AI Training Case Includes 21 Books by Korean Authors
Korea’s Literature Translation Institute has notified writers after 21 books by Korean authors appeared on a compensation list linked to a copyright settlement involving Anthropic’s AI training data.
Twenty-one books by Korean authors have been identified on a compensation list connected to a copyright settlement over training data used by U.S. artificial intelligence company Anthropic, according to Korea’s Literature Translation Institute.
The institute said it reviewed the list posted on a website handling copyright claims related to litigation against Anthropic and notified the affected writers. The books include English-language editions of works by prominent Korean authors, among them Han Kang and Hwang Sok-yong.
The development adds a Korean dimension to a wider dispute over how generative AI companies obtain books and other copyrighted material for model development. Anthropic, which operates the Claude chatbot, faced a class-action lawsuit alleging that copyrighted works had been downloaded illegally and used to train its models.
A U.S. court recognized copyright infringement involving roughly 500,000 works, and the litigation concluded in July with Anthropic agreeing to pay a total of $1.5 billion. That works out to about $3,000 per work. Korean writers who submit valid claims may therefore be eligible for compensation.
The institute’s list includes Kim Ae-ran’s “My Brilliant Life,” Kim Young-ha’s “Your Republic Is Calling You” and three other books by Kim, Kim Sum’s “One Left,” Shin Kyung-sook’s “Please Look After Mom” and two other works, three books by Han Kang including “Human Acts,” Hwang Sun-mi’s “The Hen Who Dreamed She Could Fly,” and Hwang Sok-yong’s “The Guest.”
The Literature Translation Institute said it would work with copyright-related organizations and, when necessary, actively support legal measures so that writers can receive appropriate compensation. Its intervention could help authors navigate a claims process originating in the United States and determine whether their translated editions qualify.
The disclosure comes as Korean publishing groups press AI developers for greater transparency and payment. The Korean Publishers Association issued an appeal a day earlier amid reports that overseas AI companies were seeking to purchase books in bulk for possible training use. The group argued that authors and publishers often do not know when, how or by which AI system their work has been used, leaving them unable to object or demand compensation.
The case underscores a central challenge for the AI industry: securing enough material to build capable models while respecting the rights of creators and translators. For Korea’s literary sector, it also raises practical questions about rights in translated editions and access to remedies when training data is assembled across borders.