Author

Li Xinmeng, Research Fellow, Law and Technology Institute

Abstract

In judicial practice,it still lacks a unified standard for delineating the duty of care in copyright infringement cases involving generative AI,while traditional copyright infringement frameworks also face significant challenges in this context.Drawing on system safety theory,a reasonable delineation of the duty of care across the AI value chain can help resolve the dilemmas inherent in traditional liability allocation.From the law and economics perspective,the duty of care encompasses both the level of care and the level of activity.By conceptualizing Generative AI service providers and users as potential injurers and copyright owners as victims the study explores how assigning duties of care to one or both parties influences social costs and the attainment of social welfare optimum,and identifies the key factors affecting the allocation of duties of care to one or both parties influences social costs and the attainment of social welfare optimum,and identifies the key factors affecting the allocation of duties of care.Building on this,solutions are proposed at both the legal and contractual levels.First,the article suggests adopting the "know-should have known" standard applied to internet service providers,in conjunction with dynamic systems theory,to clarify the factors shaping generative AI service providers' duty of care.Second,it emphasizes the role of contractual mechanisms—including industry self-regulation,copyright compliance,and terms of service—in allocating duties of care among stakeholders.Establishing clear duty-of-care standards can foster collaboration among copyright owners,service providers,and users,thereby addressing the challenges of liability allocation in generative AIrelated copyright infringement.

Keywords: generative artificial intelligence ; copyright infringement ; duty of care ; law and economics ; contract mechanism