The Opt-Out Mechanism to Control the Use of Copyrighted Works in AI Training:Legal and Policy Issues

Abstract

          This research aims to study the legal and policy impacts of the “opt-out” mechanism, which allows copyright holders to refuse the use of their work in training artificial intelligence (AI) through text and data mining (TDM) processes, based on the principle of informational autonomy. It also compares the legal frameworks in the European Union, Japan, and the United States.
          The article suggests that the “opt-out” mechanism is not merely a technical measure but reflects the evolution of legal norms aimed at balancing the rights of creators with the public interest in AI development. The European Union has enshrined this mechanism in its legislation, defining clear obligations for AI developers, while Japan allows extensive use of data for analytical purposes without an opt-out system. The United States, however, relies on the uncertain principle of fair use, which depends on judicial interpretation.
          Furthermore, the article highlights the emerging trend of a new structural right called the “training right” and proposes clear policy guidelines, particularly for developing countries like Thailand, to prepare for the opt-out mechanism that is becoming a new practice in the AI ​​era. This includes establishing standardized technical signals and setting up a central database system. Enacting TDM legislation and establishing international cooperation are crucial for effectively protecting copyright holders' rights while promoting responsible AI innovation.