Korean Publishers Association signs MOU to build licensed AI-training data pipeline with five data firms

What happened

On 10 November 2026, the Korean Publishers Association signed a technical-cooperation memorandum of understanding with five AI data companies — Selectstar, Alchera, AIWorks, Crowdworks and Flitto — to build a system for supplying publishing data as licensed AI training data through legal and transparent procedures.

What the agreement sets up

Under the agreement, the association will consult with publishers and act as the channel for supplying publishing data, operating the data collection and management system independently. The technology partners will provide data refinement and processing technology and support the computing infrastructure and security environments for safe handling. The stated aim is a framework under which published works can be supplied as high-quality AI training data, protecting publishers' rights while giving AI companies a stable supply of quality data.

Rights protections — and what it is not

The agreement sets out principles for protecting the rights of publishers and authors: partner firms may use the data they receive only within the scope of their business purposes, and must obtain the association's prior consent before providing or selling it to third parties. Use beyond the original purpose, such as training AI models, is prohibited without that consent, and the data and any copies must be returned or destroyed once a project ends. Rights to the works remain with the copyright holders and publication rights holders, and actual supply terms and the signing of individual contracts are premised on their consent. This is a framework MOU, not a signed licence with a named frontier lab and not a training-data licence itself.

Context

The move follows criticism of AI companies training on Korean books without permission, and follows the association's October 2026 'No Books, No AI' appeal demanding training-data disclosure and fair compensation. An association official said the group plans to use the agreement as a starting point to make publishing data a foundation of the country's AI competitiveness while building a trading system in which that value is fairly returned to publishers and authors. The five partner firms are AI data-technology companies; they are named in the body but are not yet catalog companies, so the Reviewer can link them.