“I don’t even know if I wrote my book”… The publishing industry is raising alarms over concerns about AI learning

Seoul = Yonhap News) Reporter Kim Jeong-eun = As movements to purchase large quantities of books suspected of being used for artificial intelligence (AI) learning were discovered one after another at home and abroad, concerns in the domestic publishing industry grew.

Publishing organizations such as the Korean Publishing Culture Association and the Korean Publishers Association are raising their voices one after another and seeking ways to respond. However, due to the lack of related systems, such as difficulties in verifying whether the copyright holder's book was used for AI learning, there are currently limitations in practical response.

◇ ‘Emergency’ movement to purchase books in bulk at home and abroad

The movement to purchase books in large quantities suspected of being used for AI learning was initially seen overseas, but has recently been detected in Korea as well.

Since the beginning of this year, Kyobo Bookstore has received requests to purchase one individual book from 10 overseas companies, including the United States and the United Kingdom, and about 100,000 books from each company. However, Kyobo Book Center decided not to supply it, as it judged that the possibility of it being used for AI learning could not be ruled out.

An official at Kyobo Bookstore explained, “Usually we deliver mainly to domestic libraries and the purpose is clear,” and “However, in these cases, an unclear overseas company requested delivery for unclear reasons.”

Yes24 also received a call at the end of August from a U.S. distributor wanting to purchase used books in bulk. There was one meeting regarding this, but no further progress was made since no contact was made with the company.

Yes24 said, "We are carefully watching the situation, and if there are similar inquiries, we will respond after carefully understanding the purpose."

In addition, the Literature Translation Institute of Korea recently discovered that 21 books by Korean authors were listed on the list of books subject to compensation posted on a copyright settlement site related to the lawsuit against Antropic, an American AI company, and notified the authors in question.

Previously, Antropic was embroiled in a class action lawsuit in the United States on suspicion of illegally downloading and using copyrighted work to train AI models.

In Japan, a large number of books in specialized fields such as philosophy, history, and medicine were recently observed being sold in used bookstores, raising suspicions that AI model developers were buying them to be scanned for learning purposes.

◇ Publishing organizations, appeals and fact-finding surveys… “Response by creating a council”

In relation to this, the Korean Publishing Culture Association has begun a survey of bookstores, distributors, used bookstores, publishing companies, etc. The investigation plans to ask whether there were large purchase orders suspected to be for AI learning purposes.

An official from the Korea Publishing Culture Association emphasized, "After investigating the actual situation, we will need to create a consultative body of publishing organizations to prepare response measures. Ultimately, it is important to establish a compensation system related to AI learning and legislate it."

The Korea Publishers Association recently issued an appeal calling for fair compensation when publications are used for AI learning. As of the morning of the 20th, about 200 organizations, including publishers and related organizations such as Blue Forest and Humanitas, and 570 individuals, including writers, translators, illustrators, and publishers, participated in the joint venture that began on the 15th.

Hannah Jo, CEO of Blue Forest, said, “AI technology is developing incredibly quickly, and content and books are being used for AI learning, but there is no solution for this.” She added, “The people that the publishing industry should demand copyright fees or rights from are large overseas companies, but it is a problem that cannot be won through individual fights.”

A Humanitas official also explained, "If you search on AI, you will find very detailed information, from the book introduction to the chapter summary, but they are using that information commercially without paying any fees," and added, "Publishers do not know where AI companies get the book content and how to use it."

An official from the Korea Publishers Association said, "The number of people participating in the appeal continues to increase," and added, "We plan to find a response plan in cooperation with the National Assembly and the government in the future."

◇ There is a right, but it is difficult to exercise… “Laws and systems need to be improved”

However, there are voices in the publishing industry that there are limits to response due to the current lack of legal mechanisms.

It is pointed out that a thorough system needs to be established and maintained in order for the copyright of publications to be properly protected and fair compensation provided.

Park Jeong-in, a copyright law expert and professor at Duksung Women's University's AI DynaInfo Research Institute, pointed out, "The biggest problem is that it is difficult for authors and publishers to even confirm whether their books were used for AI learning."

Professor Park emphasized, "If information on where a book was obtained and how it was used is concentrated in the company, it is difficult for the rights holder to negotiate terms of use and prove infringement even if infringement is suspected. The core of the problem is the 'structure in which rights exist but are difficult to actually exercise'."

Professor Park explained, “The most urgent measure is to institutionalize procedures that can confirm the source and usage details of learning data,” adding, “AI companies should record and preserve the data collection path and basis for use, and right holders should be able to request confirmation that their books are included.”

In addition, in order to secure practical enforcement against overseas AI companies, it was suggested that "there is a need to design the system in a way that imposes the obligation to verify learning information on businesses that provide AI models and services to the domestic market."

Jeong Jin-geun, a professor at Kangwon National University Law School, also explained, “What AI platforms use most when learning are press and book databases, and the issue is how the party who provided the materials should receive a fair compensation.” He added, “Global standards are an important issue, and we need to look at the trend of precedents in the United States.”

Professor Jeong said, "Within a year or two, the precedent in the United States will be established, and I believe there will be a move toward providing fair compensation. In this case, a new system for collecting, distributing, and calculating compensation must be created, and related legal systems must be reorganized. In this process, administrative leadership led by the state is also needed."