Anthropic to Compensate for Illegally Downloaded Korean Books

U.S. artificial intelligence (AI) company Anthropic has been found to have illegally downloaded English versions of books by Korean authors, including Kim Ae-ran, Han Kang, and Hwang Sok-yong, and used them to train its large language model (LLM) Claude.

The Literature Translation Institute of Korea stated on October 16, “We identified that 21 books by Korean authors are listed among the 482,460 cases of compensation-eligible works on the settlement site related to the Anthropic lawsuit and have already notified the authors.” On the same day, the Korean Publishers Association began collecting signatures for an appeal titled “South Korean Appeal – No Books, No AI” opposing AI companies’ mass acquisition of books, marking the domestic publishing industry’s active response to what it calls “knowledge theft” by AI firms.

According to the institute, the Korean authors whose works were illegally downloaded by Anthropic include Kim Ae-ran, Kim Young-ha, Kim Sum, Kim Hye-soon, Bae Su-ah, Shin Kyung-sook, Jung You-jeong, Choi Yoon, Han Yu-joo, Hwang Seon-mi, and Hwang Sok-yong (in alphabetical order). Anthropic currently plans to compensate approximately 3,000 dollars (approximately 4 million Korean won) per work for illegally acquired content, following a U.S. court ruling. Korean authors are also expected to receive compensation if they participate in the claims process.

Domestic authors have openly expressed discomfort over AI companies’ unauthorized use of their works. Kim Ae-ran stated, “This is not merely an issue of individual intellectual property infringement but feels like an act of specific capital and power undermining humanity’s shared assets and public values,” adding, “It appears to be a process of razing the forest of books under the guise of development and devastating our intellectual soil and publishing environment.” Kim Sum remarked, “It is overwhelming for an individual to confront something that transcends borders,” calling the situation “a battle against a colossal, ghost-like opponent.”

In its appeal, the Publishers Association pointed out, “Generative AI develops through books, yet the creators who made those books cannot even know when, how, or which AI their works are being used in,” continuing, “Without knowledge, they cannot ask questions, refuse, or demand fair compensation.” It emphasized, “We do not reject technology,” and questioned, “If technology that grew on today’s books blocks the path for tomorrow’s books to emerge, can that development truly last?” The association demanded the government, National Assembly, and AI companies to ensure transparency in training data, establish a principle of obtaining rights holders’ permission, and provide fair compensation. It also expressed solidarity with the ‘Frankfurt Appeal’ announced by the German Publishers and Booksellers Association and over 120 publishers at the Frankfurt Book Fair on October 6.

As of 5 p.m. on October 16, 173 publishing companies and organizations, including Moonji Publishing and Park Youngsa, along with 355 individuals, had joined the appeal.