Google is negotiating book purchases with South Korean publishers for AI training, according to publishing-industry sources. The reported talks raise questions about authors’ consent and smaller publishers’ bargaining power. Contract parties and terms remain unconfirmed, while proposals requiring disclosure of training-data sources are still pending in South Korea’s legislature.
Publishing-industry sources told Seoul Economic Daily on October 18 that Google wants to purchase books for AI training and is discussing deals with leading publishers in South Korea. Several sources understood that some contracts had been completed, but the publishers involved and the terms could not be confirmed. The newspaper received no response to its questions from Google Korea. The talks reportedly began in the first half of the year.[1]
The negotiations follow controversy over Anthropic’s bulk purchase of Korean books from a distributor. Google’s approach to publishers still raises a question about whether those publishers have obtained authors’ individual consent for AI training. An industry official warned that small publishers face difficulties meeting requested file formats and quality standards, as well as judging compensation. The official said financial pressure on one publisher could establish a low reference price for other negotiations.[1]
The debate surrounding Google’s talks distinguishes piracy damages from permission to train an AI model. Industry discussions include varying license terms by the scope of use, recurring payments when answers refer to books, and how proceeds are divided between authors and publishers. South Korea’s National Assembly has yet to pass proposed AI Framework Act changes covering transparency about training material and a procedure for rights holders to inspect it.[1]
