Anthropic faces copyright dispute over Korean authors' books used to train Claude
Anthropic used English translations of works by Korean writers such as Han Kang and Hwang Sok-yong to train its Claude model without permission, prompting a settlement and calls for compensation.
According to the Literature Translation Institute of Korea, Anthropic incorporated English translations of 21 Korean books into the training data for its Claude AI system without obtaining author consent. The list features prominent writers such as Han Kang, Hwang Sok-yong, Kim Young-ha and Shin Kyung-sook. Anthropic had earlier been sued in the United States over alleged illegal downloading of copyrighted works and reached a settlement in July, committing to a payment whose amount was not disclosed.
The Korean Publishers Society (KOPUS) has called for equitable compensation, emphasizing that AI models depend on literary content and that authors are often left unaware of how their works are used. The institute notified the affected authors in March, highlighting growing concerns over cross-border AI training practices.
