Abstract

As the use of TDM and AI expands, clauses restricting the use of content for TDM and AI are increasingly being included in license agreements. While such clauses have the potential to raise significant issues for academic research and library operations, related discussions remain limited in South Korea. In response, this study conducted an empirical analysis of the restrictions on TDM and AI usage in license agreements. Through an inductive approach, the study developed a classification framework for TDM and AI clauses based on their actual expressions in licensing documents. The clauses were categorized into four domains: TDM input, TDM output, use of AI tools, and use for training of AI. Separate classification criteria were established for (1) TDM input, use of AI tools, and AI training, and (2) TDM output. The resulting framework was applied to examine the current state of each clause type. Across all types, the most frequently observed category was “conditionally permitted (public),” though differences were observed in the levels of permission granted. The classification criteria for TDM and AI clauses are expected to serve as a consistent analytical tool for tracking changes as such clauses continue to evolve.