생성형 AI 시대 저작권 보호를 위한 학습 불가능한 예제와 기술적 보호조치의 개선 방안 연구

초록

This study analyzes the institutional and legal gaps that arise in the Copyright Act when generative AI systems learn from copyrighted works and other data without explicit authorization. Comprehensive measures are proposed to address these deficiencies. By conducting a comparative analysis on the copyright laws of major jurisdictions, including text and data mining (TDM) exceptions and recent case law, the study identifies critical limitations within the current domestic legal framework in regulating AI training practices. Furthermore, through empirical validation, the research examines the feasibility of the unlearnable examples technique as a practical protection measure to safeguard copyrighted materials during the AI learning process. Building on these insights, the study ultimately proposes complementary institutional and policy reforms designed to establish a balanced ecosystem that ensures both effective copyright protection and sustainable advancement of generative AI innovation.

키워드