Author

Wu Yaxi, Phd student, Renmin Law School 

Abstract

The rapid iteration of generative artificial intelligence has reshaped the production and dis- semination of information in the Web 3. 0 era. Traditional platform tort liability rules, centered on the safe harbor rule and supplemented by secondary liability, have gradually become ineffective. In the context of generative AI platforms, copyright infringement has evolved from static, isolated individual cases into dy- namic, systematic structural risks. The traditional liability regime, which focuses on actual knowledge and adopts case-specific remedies, is ill-suited to the characteristics of generative AI, where tort risks are embedded in models and infringing content is continuously and dynamically generated. By sorting out the elements of the traditional liability framework and analyzing the evolution of generative AI - related torts, this article argues that the tort liability paradigm should shift from case-specific remedies to system- ic risk governance. It proposes moderately expanding the knowledge requirement, drawing on the doctrine of willful blindness, and establishing a risk identification standard based on repeatability, scalability, for- eseeability, and controllability. On this basis, the existence of deliberate avoidance by AI platforms can be determined to establish tort liability. Remedial mechanisms should extend from the content level to the model level, prioritizing risk prevention, and adopt punitive damages to incentivize generative AI plat- forms to participate in copyright governance.

Keywords: Generative AI, Copyright Infringement, Platform Liability, the Safe Harbor Doctrine, Systemic Governance