What happened
South Korean publishers and bookstore chains are investigating bulk purchase inquiries from overseas companies suspected of acquiring books for AI training. The Korean Publishers Association has launched a survey to identify these transactions and is working toward establishing compensation arrangements and legislative protections.
According to the South China Morning Post, South Korean publishing groups are investigating suspicious bulk book orders that they suspect are linked to artificial intelligence training. Kyobo Book Centre, a major bookstore chain, reported that 10 overseas companies, including businesses in the United States and Britain, approached it since early this year. Each company requested only one copy of approximately 100,000 titles. A Kyobo representative stated that many buyers introduced themselves as book wholesalers, but some identified data-related companies as the final recipients. The bookstore declined the requests because it could not rule out AI training as the purpose.
Yes24, another bookstore, reportedly received an inquiry from a US distributor in late August regarding the bulk purchase of used books. Although the two parties held one meeting, discussions ceased after the distributor stopped responding. Additionally, the Literature Translation Institute of South Korea notified local authors after identifying 21 of their books on a compensation eligibility list associated with a US copyright settlement involving Anthropic. This settlement stems from a class-action lawsuit alleging that Anthropic illegally downloaded copyrighted works for model training.
The Korean Publishers Association began surveying bookstores, distributors, second-hand booksellers, and publishers last week to identify potentially AI-related bulk orders. An association official indicated that publishing organizations should form a joint body following the survey to establish compensation arrangements and push for legislation. The Korean Publishers Society also issued an appeal demanding fair compensation when publications are used for AI training. As of Wednesday, over 240 publishers and organizations, along with 702 authors, translators, and illustrators, had joined this appeal.
Industry leaders argue that individual rights holders cannot effectively confront large overseas technology companies alone. Cho Han-na, head of publisher Prunsoop, emphasized the difficulty of negotiating without information about where AI companies obtain book content and how they use it. Park Jung-in, a copyright specialist at Duksung Women’s University, noted that information gaps make it difficult to prove suspected infringement or negotiate licensing terms, describing the core problem as a structure where rights exist but are difficult to exercise in practice.
Why it matters
This development highlights a growing global friction point between AI developers and content creators regarding data acquisition. By organizing a coordinated response, South Korean publishers are attempting to address the opacity of AI training data sources, which currently prevents rights holders from verifying usage or negotiating fair licensing terms. This could set a precedent for how other regions handle unauthorized bulk data acquisition for machine learning.
The opacity of AI training data acquisition is a significant barrier for content creators worldwide. South Korean publishers are attempting to bridge this gap by coordinating their efforts to identify and document bulk purchases that may be intended for AI training. This organized response aims to create leverage for negotiating compensation and influencing legislation.
The situation underscores the legal and ethical complexities of using copyrighted material for AI development. While AI companies often claim fair use or rely on public data, publishers argue that bulk, anonymous purchases are a method to bypass licensing agreements. The involvement of overseas companies highlights the transnational nature of this issue, requiring international cooperation or robust domestic laws to protect local creators.
The reference to the Anthropic settlement in the US provides context for the potential legal outcomes of such disputes. If South Korean publishers can establish a clear link between bulk purchases and AI training, it may strengthen their position in seeking compensation or regulatory action. This could lead to new standards for and transparency in the AI industry.
Interactive Mechanism: How It Actually Works
Explore the underlying technology behind this development interactively.
crm_get_transaction(id='4092').Why can ethical evaluation not be reduced to one model score?
What to watch next
Watch for the results of the Korean Publishers Association's survey, the formation of a joint industry body, and any subsequent legislative proposals in South Korea regarding AI training data transparency and compensation.
The outcome of the Korean Publishers Association's survey will be crucial in determining the scale of the issue and the specific companies involved. This data will likely inform the formation of the proposed joint body and the specific demands made to legislators.
Legislative developments in South Korea regarding AI training data transparency and compensation will be a key indicator of the government's response to the industry's concerns. New laws could require AI companies to disclose their data sources or establish a licensing framework for copyrighted works.
Reactions from AI companies and international publishers will also be important. If other countries face similar issues, South Korea's approach could serve as a model for coordinated industry responses to AI data acquisition practices.