Intermediary Liability for IP Infringement in the Age of AI: Duty of Care and Institutional Design

Abstract

The rapid rise of artificial intelligence (AI) has transformed the dynamics of intellectual property (IP) enforcement, positioning intermediaries not merely as passive conduits but as active participants in content creation, training, and dissemination. Traditional liability regimes, such as safe harbor provisions, are increasingly inadequate to address the scale and complexity of generative AI. Yet existing scholarship remains fragmented, with debates focusing narrowly on copyright subsistence or platform immunity while neglecting the institutional role of intermediaries in shaping infringement outcomes. To address this gap, the study employs a qualitative, multi-method approach that integrates doctrinal and policy analysis, cross-jurisdictional comparison, and case studies of recent disputes in the United Kingdom, United States, and China. Findings demonstrate that intermediaries operate as co-creators and gatekeepers, requiring a layered duty of care that differentiates responsibilities across dataset construction, model training, and content dissemination. This framework moves beyond binary liability models and clarifies how institutional design influences efficiency, legitimacy, and adaptability. The research contributes a theoretically grounded and policy-relevant model for calibrating intermediary liability, offering guidance for regulators, platforms, and rightsholders, while pointing to the need for transnational coordination in governing AI-driven IP ecosystems.