Works by Han Kang and Hwang Sok-yong also included in Anthropic copyright lawsuit ... 21 books by Korean authors eligible for compensation

사진 확대

Anthropic, a U.S. artificial intelligence (AI) company, has agreed to pay a substantial settlement in the United States over its use in developing AI models of books obtained without copyright holders’ permission. Works by Korean authors, including Han Kang and Hwang Sok-yong, have also been confirmed as covered by the settlement. With 21 books by Korean authors listed among those eligible for compensation in the U.S. copyright class-action lawsuit, attention is also turning to copyright protection and compensation for creators in South Korea.

On the 17th, the Literature Translation Institute of Korea said it had confirmed on a website related to the settlement of the U.S. copyright class-action lawsuit against Anthropic that 21 books by Korean authors were eligible for compensation. The institute notified the authors and provided information on options for seeking redress and claiming compensation.

The books identified by the institute as eligible included Human Acts by Nobel Prize in Literature laureate Han Kang, The Guest by Hwang Sok-yong, Please Look After Mom by Shin Kyung-sook, The Empire of Light by Kim Young-ha, My Brilliant Life by Kim Ae-ran, One by Kim Sum, and The Hen Who Dreamed She Could Fly by Hwang Sun-mi. Three of Han Kang’s works, including Human Acts, were identified, as were four of Kim Young-ha’s, including The Empire of Light, and three of Shin Kyung-sook’s, including Please Look After Mom.

These works, including books translated into English and published overseas, were included in a list of materials that Anthropic is said to have obtained through illegal downloads.

Anthropic became embroiled in a U.S. class-action lawsuit over allegations that it downloaded illicit e-book materials from the internet without copyright holders’ permission and used them to develop AI models. A key issue in the case was whether Anthropic’s acquisition and retention of large volumes of illicit e-book materials was lawful.

A U.S. court ruled that Anthropic’s use of copyrighted books to train AI models constituted fair use. However, it found that downloading books from piracy sites and keeping them separately amounted to copyright infringement.

Anthropic and the plaintiffs later reached a settlement totaling $1.5 billion (about 2.1 trillion won), which a U.S. federal court gave final approval to last July. Anthropic did not admit legal liability as part of the settlement. The settlement covers about 500,000 titles and is considered a major compensation case in AI copyright disputes.

Based on the total settlement amount, compensation is estimated at about $3,000 (about 4.3 million won) per work. Actual payments may vary, however, depending on litigation costs and how the funds are divided among copyright holders. According to documents submitted to a U.S. court last September, the estimated initial distribution was about $2,204 per work, with the remaining settlement funds expected to be paid out later.

The Literature Translation Institute of Korea plans to help Korean authors exercise their rights. An institute official said, “We will work with relevant copyright organizations and actively support legal action and other measures as needed so that authors receive fair compensation.”

Calls are growing within South Korea’s publishing industry for greater transparency and a compensation system governing AI companies’ use of copyrighted works.

The Korean Publishers Association said it had issued an “Appeal to Korea” on the 16th, calling on the government, the National Assembly and AI companies to disclose the sources of AI training data, obtain rights holders’ consent and pay fair compensation. It also said it had begun gathering signatories. The move followed the detection of efforts by overseas AI companies to purchase large quantities of books believed to be intended for training.

In the appeal, the Korean Publishers Association said, “Without books, there is no AI,” and pointed to how difficult it is for creators even to find out which AI models are using their works. It went on to call for transparent disclosure of the sources and usage records of works used to train AI, and for a compensation system that protects creators’ rights.

This article has been translated by GripLabs Mingo AI.