[AI Revolution](222) In an Era of Rampant AI Book Theft... Mentat: "Supporting Legal Transactions"

Interview with Mentat CEO Jeon Byeonguk
Working with around 90 publishers... 34,000 titles secured
Taking responsibility from obtaining authors' consent to quality verification
Targeting 20 billion won in cumulative AI licensing contracts

"A market in which artificial intelligence (AI) becomes a reader will emerge."

Jeon Byeonguk, CEO of Mentat (pictured), identified AI as a new source of demand for publications in an interview with The Asia Business Daily on November 15. His plan is to expand the publishing industry's revenue base by supplying books read by people as data that AI can learn from and use to generate answers. "With the publishing market facing difficulties and the AI industry growing rapidly, I wanted to find new business opportunities," Jeon said, explaining what led him to start the company.

Mentat CEO Jeon Byeonguk speaks in an interview with The Asia Business Daily on the 13th. Photo by Lee Eunseo.

Last month, it came to light that Anthropic had obtained works by Korean authors without authorization and purchased large quantities of books in Korea, bringing copyright holders' consent and fair compensation for AI training data to the fore as key issues. In the publishing industry, there is growing concern that purchasing a book and obtaining the right to use it for AI training are separate matters.

In response, Mentat has set out to facilitate legal data transactions between publishers and AI companies and build an ecosystem in which copyright holders receive fair compensation. Mentat works with publishers to secure book licenses and supply the books to AI companies. It tags information such as titles, authors, and publishers as metadata, making it easy to find the materials needed.

Mentat currently works with around 90 publishers and has secured 34,000 titles, including books and magazines. "Publishers can earn new revenue beyond traditional book sales, while AI companies can use data that is difficult to obtain, provided they pay a fair price," Jeon said. "The key is that we supply books with clearly established rights."

He cited books' long-form context and editorial process as sources of value in book data. Unlike fragmentary information that can be gathered online, books organize knowledge systematically according to their tables of contents, he explained. "These are texts whose facts have been verified by professional editors and that have undergone proofreading, copyediting, and editing," he said. "At present, demand is high for nonfiction that covers facts and knowledge, such as specialized science and engineering books." He added, "We also review books intended for AI use to manage data quality."

Jeon's more than 10 years of experience in the publishing industry has given him a foundation for working with publishers. He served as head of the digital business team at Changbi, where he gained experience in e-book and audiobook businesses, and led new ventures such as audio dramas and chat fiction at Kakao Entertainment. His deep understanding of publishers' contract structures and business processes is an advantage, as is his ability to consider how to obtain authors' consent when securing licenses.

The cumulative value of Mentat's AI licensing contracts with global Big Tech companies currently stands at approximately 15 billion won. Its goal is to exceed 20 billion won by the end of 2026. The company's business currently centers on supplying data for AI model training, but it plans to expand into reference and citation licensing in the future. Under this model, AI will find and use relevant content from books to answer users' questions, with fees charged according to usage. To support this, Mentat plans to secure 100,000 titles by 2027.

Mentat is preparing to enter the data market for AI agents in particular. This includes data that helps agents produce accurate results when they search multiple documents, synthesize their contents, or perform tasks. The company is also considering entering the field of "reinforcement learning environments," where agents perform tasks, have their results evaluated, and learn. "The data AI needs will change as technology evolves," Jeon said. "We started with books, but we aim to build a business that connects AI and data beyond publishing."

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