Survey of 188 Publishers Shows 20.7% Have Published AI-Utilized Works
Proposal for Three-Tier Classification: Human Authored, AI-Assisted, and AI-Generated Publications
Authors Must Also Disclose AI Usage Scope; 12 Publishing Organizations Annou

It has been revealed that three out of four domestic publishing companies are already utilizing generative artificial intelligence (AI) in their content production processes. In addition, one out of five publishers reported having actually published books making use of AI. With AI rapidly permeating the publishing field, industry officials have proposed an “AI Publication Labeling System” to inform readers of the extent to which AI was involved in the making of a book.

On the 29th, publishing industry officials announced the "Principles for Use of AI by Publishers and Authors for Readers" at the Publishing and AI Policy Forum held at Mapo Central Library in Mapo-gu, Seoul. (From left in the photo) Seongjin Kim, critic at Literary Critique Collective; Hanbyul Hong, translator; Gyeongdong Song, secretary general of the Korean Writers' Association; Youngwan Hong, chairman of the Korea Publishers Association. Korea Publishers Association Photo by Korea Publishers Association

On the 29th, publishing industry officials announced the "Principles for Use of AI by Publishers and Authors for Readers" at the Publishing and AI Policy Forum held at Mapo Central Library in Mapo-gu, Seoul. (From left in the photo) Seongjin Kim, critic at Literary Critique Collective; Hanbyul Hong, translator; Gyeongdong Song, secretary general of the Korean Writers' Association; Youngwan Hong, chairman of the Korea Publishers Association. Korea Publishers Association Photo by Korea Publishers Association

View original image

On the 29th, the Korea Publishers Association held a publishing and AI policy forum themed “From Crisis to Coexistence: The Future Path of the Publishing Ecosystem in the AI Era” at Mapo Central Library in Mapo-gu, Seoul. This was a follow-up to an emergency forum convened in response to controversy over so-called “click-to-publish” books, which are produced rapidly using generative AI—a debate that began last April. During the event, the results of research and discussions carried out over the past five months were shared, presenting criteria for distinguishing AI-assisted publications and unveiling the “Principles for Use of AI by Publishers and Authors for Readers.”


AI is already firmly embedded in the publishing landscape. According to Jeongin Park, professor at Duksung Women’s University and head of the DynaInfo AI Research Lab, a survey conducted from July 1st to 31st covering 188 domestic publishers found that 76.1% had either introduced generative AI on a trial basis or were using it for some or many of their work processes. Also, 20.7% responded that they had actually published works utilizing generative AI. Furthermore, 85.1% of companies stated that it was necessary, or at least partially necessary, to have an identification system to distinguish AI-generated publications.


The recognition of what constitutes an “AI-generated book” varied greatly depending on the degree of human involvement. When AI generated the entire text, 88.3% of respondents considered it to be an AI-authored work. However, when the entire text created by AI was subsequently reviewed by a human, the percentage dropped to 79.8%, and when AI only translated the text, this figure fell to 50.5%. If AI drafted only part of the manuscript and a human finalized it, only 34.6% recognized it as an AI publication. Ultimately, rather than simply how many sentences were written by AI, the key question lay in how much humans controlled and verified the output.


This is reflected in the criteria proposed by the publishing industry, which suggest dividing publications into three groups: 1) human-authored works, 2) AI-assisted works, and 3) AI-generated publications. “Human-authored works” are those in which humans are the main creative agents, including cases where AI is only used for simple editing, proofreading, or drafting outlines and tables of contents. It is proposed that such works do not require a separate AI label.


“AI-assisted works” refer to cases where content created by humans is supplemented with AI, or where content generated by AI has been thoroughly controlled and verified by humans before completion. In contrast, “AI-generated publications” are those in which AI produced the content with insufficient human control or verification. The core of the proposed system is that for categories 2 and 3, publishers should disclose the fact that AI was used and to what degree humans controlled and reviewed the material. The primary judgment criterion should not be based on the “proportion” of AI-generated sentences but rather whether a human is ultimately capable of taking responsibility for the finished work.


The responsibilities of authors will also increase. According to the proposal, authors are expected to declare which AI tools were used and how extensively AI was employed in the core content when submitting a manuscript to the publisher. Publishers and authors should agree to the extent of AI use and specify this in their publishing contracts. The principles also stipulate that, if a publisher intends to provide a book as AI training data, the rights holder—such as the author—must be informed and consent must be obtained so that fair compensation may be ensured.


On the 29th, at Mapo Central Library in Mapo-gu, Seoul, stakeholders from publishing, authorship, and distribution sectors are discussing AI usage standards and response strategies for the publishing ecosystem during the policy forum titled "From Crisis to Coexistence: The Future Path of the Publishing Ecosystem in the AI Era." Korea Publishers Association

On the 29th, at Mapo Central Library in Mapo-gu, Seoul, stakeholders from publishing, authorship, and distribution sectors are discussing AI usage standards and response strategies for the publishing ecosystem during the policy forum titled "From Crisis to Coexistence: The Future Path of the Publishing Ecosystem in the AI Era." Korea Publishers Association

View original image

The labeling system aims not only to secure readers’ right to know but also to protect the rights of a publication as “training data.” By clarifying the rights relationships of works where human creativity and verification were involved (categories 1 and 2), the industry hopes to establish a basis for negotiating contracts or compensation with AI companies regarding the use of learning data in the future. Professor Jeongkeun Jeong at Kangwon National University School of Law suggested that since it is difficult for individual writers or publishers to secure compensation for the use of their works in AI training, publishers should create a structure to negotiate collectively.


However, questions remain, such as whether “simple labeling makes a book trustworthy.” Youngho Ryu, head of publishing at Kyobo Book Centre, pointed out that simply disclosing the use of AI does not guarantee the credibility of the content, and suggested the need for a verification and distribution system where publishers share review information with distributors, and bookstores communicate this information to readers. In the field of translation, concerns were raised about whether the proposed three-stage classification is appropriate. Hanbyul Hong, a translator, emphasized that it is difficult to assess AI usage in translation simply by proportion, and argued that it is more important to clarify what processes utilized AI and whether the translator takes responsibility for the outcome.


The “Principles for Use of AI by Publishers and Authors for Readers,” released on this day, consist of 5 chapters and 20 articles. The key principle is that AI is a tool only and cannot hold authorship status, and that all responsibility for the creation and its outcomes lies with human beings. Twelve organizations related to publishing and authorship, including the Korea Publishers Association, Korean Writers' Association, and the Association of Humanities and Social Science Publishers, participated in the preparation of these principles. The Korea Publishers Association stated that this marks the world’s first instance of publishers collectively adopting documented AI usage principles.


However, these remain recommendations without legal force. Details still need to be refined, such as how much AI usage a publisher can realistically verify about an author, or who will decide ambiguous cases regarding categorization. The Korea Publishers Culture Industry Promotion Agency, a major publishers’ association, did not participate in these principles. While the publishing industry has set its first standard by advocating for transparency—clearly disclosing the extent of AI use and responsibility, rather than prohibiting AI outright—further discussion is required before these can be established as an industry-wide rule.



Youngwan Hong, president of the Korea Publishers Association, emphasized that these principles are not intended to place a “scarlet letter” on AI usage. He stated, “Our goal is to create an environment in which publishing and AI can coexist, and readers can feel assured when choosing books.”


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