Anthropic settlement list includes at least 22 works by Korean authors

Chosunbiz reported that at least 22 works by Korean authors were among the 482,460 works covered by the settlement.[3] The list includes three books by Han Kang—“Human Acts,” “The Vegetarian” and “The White Book”—as well as four works by Kim Young-ha and three by Shin Kyung-sook.[3] It also includes works by Hwang Sok-yong and Kim Ae-ran.[3] Separately, the Literature Translation Institute of Korea said it found 21 Korean-authored books on the settlement website and notified their authors.[2] These reported counts refer to different descriptions of the list; the evidence does not explain the difference.[2][3]

The U.S. court found that using books to train AI models is fair use under the Copyright Act, but did not recognize downloading and storing pirated copies to obtain training books as fair use.[3] The settlement followed Anthropic’s acquisition of more than seven million digital copies from illegal book-sharing sites in 2021 and 2022.[3] The report says Anthropic obtained at least five million books from LibGen in 2021 and at least two million from Pirate Library Mirror in 2022.[3] The distinction matters: the court’s finding on training does not also establish that obtaining pirated copies was fair use.[3]

Anthropic agreed to pay $1.5 billion for approximately 500,000 pirated books, and the settlement provides about $3,000 in damages per work.[3][2] Korean authors can seek compensation by participating in the claims process.[1] The evidence does not specify how an author files a claim, what documentation is required or when claims must be submitted. The Literature Translation Institute of Korea said it would support legal action if necessary to help authors receive fair compensation.[1] Authors and publishers considering next steps should therefore distinguish the stated per-work damages from any individual payment, which the evidence does not detail.[2][1]

The Korean Publishers Association appealed for transparency about AI companies’ use of copyrighted works and fair compensation, saying creators do not know when, how or which AI used their works.[1] A separate publishing-industry appeal called for transparent data use, permission from rights holders and fair compensation.[4] It had been signed by 163 publishing companies and organizations and 333 writers, translators and publishing professionals.[4] Meanwhile, Anthropic has obtained training materials since 2024 by buying print books in bulk and digitizing them; Woongjin Book Centre was reportedly included as a partner.[3]