Who Announced the Principles and Where
The principles were announced at "Beyond Crisis Toward Coexistence: The Way Forward for the Publishing Ecosystem in the AI Era", a policy forum held by the Korean Publishers Society at Mapo Central Library in Seoul[1][2]. According to Yonhap News Agency, they were developed through a survey of publishers, expert review, and several rounds of discussion. They also represent a commitment the publishing industry made on its own before the relevant institutional framework was put in place[1].
Hong Young-wan, chair of the Korean Publishers Society, said the principles were "not intended to reject AI technology or brand AI-generated material with a scarlet letter"[1].
The Core Principles Put People in Charge and Treat AI as a Tool
The reports confirm four core points in the principles[1][2].
- People create books and bear final responsibility. AI is only a tool and has no status as an author or coauthor.
- If AI is used, disclose that fact and the extent of its use in a way readers and users can recognize.
- Using publications to train AI requires the rights holder's consent and fair compensation.
- Set out recommendations for drafting publishing contracts and standards for using a publisher's own books as training data.
The Three Categories and Their Labels
The principles divide publications into three categories based on whether they have been verified by a human: "human-authored work", "AI-assisted work", and "AI-generated publication". They recommend labeling "AI-assisted work" and "AI-generated publication" as such[1][2]. In other words, books classified as "human-authored work" are not covered by the labeling recommendation.
The specific criteria separating the three categories are difficult to establish from these reports alone. Once the full text of the principles is released, this is the first part authors should examine.
What Authors Need to Do: Tell the Publisher in Advance
The provision that directly concerns authors is disclosure. The principles call for authors to tell publishers in advance at which stages and to what extent they used AI[1][2]. Unless you record which tools you used, at what stages, and how extensively while working on your first draft, you will have to rely on memory when the publisher asks later. If these principles take hold in publishing practice, records of AI use will become material to hand over with the manuscript.
Keep the record on a single page as you write. Four fields are enough: tool, stage of use, scope of use, and what a person verified. Because the three categories depend on whether a human verified the work, don't leave the last field blank.
Where the Industry Stands
In the forum's keynote presentation, Professor Park Jeong-in of the AI Dynainfo Research Institute at Duksung Women's University presented the results of a survey of 188 Korean publishers conducted in July[1].
Professor Park concluded that AI was already routinely used for assistance, but its use to generate core content remained limited[1].
- 9.6% said they actively used generative AI, 45.7% used it for some tasks, and 20.7% had introduced it on a trial basis.
- 20.7% had published at least one title whose main content was created by generative AI.
- Asked when an AI label should be required, the most common response, at 64.9%, was "when AI generated a substantial portion of the main text".
- 66% agreed that contracts needed clauses on AI disclosure and liability in disputes but said they had not yet made any preparations.
What Remains Unresolved
These principles are recommendations from some publishing and authors' organizations, so there are no sanctions for not following them. Seoul Shinmun noted that they are not mandatory and that there is still a long way to go before they become institutionalized[2]. Professor Jeong Jin-geun of Kangwon National University School of Law, who joined the panel, suggested designing guidelines to encourage responsible AI use rather than imposing uniform restrictions on it[1].
References
- [1] Yonhap News Agency, report dated September 29, 2026.
- [2] Seoul Shinmun, report dated September 29, 2026.
Frequently asked questions
Will I be penalized if I don't follow the principles?
These are voluntary principles with no legal force, so there are no sanctions for not following them. However, they include recommendations to consult when drafting publishing contracts. Authors entering into agreements with publishers are likely to encounter this standard when drawing up a contract or submitting a manuscript.
If AI wrote much of the manuscript, should I list it as a coauthor?
No. The principles treat AI as a tool and do not recognize it as an author or coauthor. People create books and bear final responsibility. Instead, the principles call for disclosing to readers that AI was used and to what extent.
How much AI use requires a label?
The specific criteria separating the three categories could not be confirmed from the reports. Once the full text of the principles is released, this is the first part to check. Until then, publishers' views can serve as a reference. In a July 2026 survey of 188 publishers, the most common response, at 64.9%, was that labeling should be required "when AI generated a substantial portion of the main text".