Abstract: The rapid advancement of generative artificial intelligence (GAI) presents significant structural tensions with existing legal frameworks governing tort liability for service providers. This study identifies four critical dimensions of these challenges. First, regarding the doctrine of liability attribution, the application of strict liability reveals a tripartite jurisprudential contradiction. Second, in terms of legal status, GAI service providers exhibit a dual nature—combining technical architectural neutrality with active content generation—which renders traditional classifications of network service providers insufficient. Third, concerning the duty of care, the "algorithmic black box" and dynamic self-evolution of models give rise to a dilemma between excessive defensiveness and liability evasion. Fourth, in terms of rule application, the distributed content-generation mechanism leads to systemic failures in the traditional "notice-and-takedown" procedure. In response, this paper proposes a new regulatory paradigm: establishing the principle of fault liability as the core standard in infringement adjudication; reclassifying GAI service providers as "novel network service providers" to reflect their hybrid legal character; constructing a dual-track duty of care system that combines passive responses with proactive prevention; and transforming the "notice-and-takedown" procedure into a "process governance-dynamic response-exceptional balancing" framework. These measures aim to strike a dynamic balance between fostering technological innovation and safeguarding rights.