Copyrights in Publishing Become AI Training Data
A New Marketplace for Publishing Copyrights: Without Protective Barriers, It's a Race to the Bottom
The Publishing Industry's Fear of Undervaluation Looms Ahead of the "Declaration of Humanity"

The theme of this year's Seoul International Book Fair is "Declaration of Humanity." Declarations are made in times of uncertainty. If being human were self-evident, there would be no need to declare it. During this celebration of books, the questions facing the publishing industry have grown even colder. Do only humans still read books? Or will people eventually buy books that machines have already read first?


Citizens visiting the '2025 Seoul International Book Fair' held at COEX in Gangnam-gu, Seoul, are reading books. Photo by Yonhap News Agency

Citizens visiting the '2025 Seoul International Book Fair' held at COEX in Gangnam-gu, Seoul, are reading books. Photo by Yonhap News Agency

View original image

There is a saying circulating in the publishing industry: If people are not going to read, shouldn't at least AI do it? It was meant as a joke, but it wasn't funny. New releases no longer last long on bookstore shelves. They go to storage, end up on return lists, and at some point, the cost of shredding is calculated. When someone says, "I wish AI would at least make it a hot topic by learning from it," what is more evident than anger over copyright infringement is the resignation that the book will be forgotten.


The American publishing industry has already taken the issue to court. Platforms like LibGen, Z-Library, Anna's Archive, and WeLib, once gateways for illegal reading, are now being identified as warehouses for AI training data. While a single pirated book may take away one sale, a book used as training data becomes material for countless summaries, translations, and answers. Books are transformed from finished products into raw materials, and those materials are then used to target the human reading market once again.


The Hollywood writers' strike raised a similar question first: When AI writes or edits a script, what is the value of human writers' labor? The publishing industry's question is related but distinct. When AI reads and learns from a book, what is the value of that book? One is about the price of creative labor, while the other is about the price of creative data.


This is not a distant issue for Korea either. Booktokki and Newtokki have been eating away at creators' revenues in the web novel and webtoon markets for years. On one side, there has long been a market for pirated reading, and on the other, "one-click publishing" has surged. Books are not selling, yet text that looks like a book is easier than ever to produce. The Korean publishing industry's anxiety is twofold: the fear of one's book being fed into a machine, and the fear that books produced by machines will flood the market once again.


The law protects specific forms of expression, but it does not protect style or ideas themselves. This is a necessary principle. However, AI feeds on that gap. It statistically absorbs the rhythm between sentences, the conventions of genres, and the order of explanations. The law looks at expressions; the model consumes patterns. What the publishing industry fears is not a few copies being duplicated, but books being invisibly dismantled and used to replace the book market itself.


The "Publishing AI Data HUB" proposed by the Korean Publishers Association is an attempt to turn this anxiety into a transaction. The idea is to collect publishing content, sell it to AI companies, and return the profits to publishers and authors. The direction is right. However, simply creating a marketplace does not guarantee fair pricing. Big tech companies can afford to wait, but small publishers cannot. Next month's printing costs, storage fees, and royalties are billed first. The side that can wait holds the pricing power.


Therefore, what is needed before a hub is a fence. The European AI Act has mandated the disclosure of summaries of training content, and The New York Times and American writers have taken OpenAI to court over the sources and substitution effects of training data. Korea, too, cannot open a data marketplace without prior systems for usage disclosure, author consent, standard contracts, minimum prices, and collective bargaining mechanisms. The moment someone signs a bargain-basement contract, that price becomes the market standard. "Reasonable pricing" is only possible when both sides have similar negotiating power.



If the "Declaration of Humanity" at the Seoul International Book Fair is not to ring hollow, we must now ask the price when machines read books written by humans. If the publishing industry cannot set its own price tag, big tech will do it for them. At that point, the price paid for books read by machines will not be a technology company's cost, but the survival cost for the smallest publishers and the most powerless authors.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing