Author Information

Zhang Xinbao, Wu Yuzhang Chair Professor, Renmin Law School

Bian Long, Ph.D. Candidate, Renmin Law School

Abstract

Generative artificial intelligence doesn't conform to the concept of product from the perspective of the distinction between products and services. Its main function is to output specific content and information, and service providers shouldn't bear product liability. Generative artificial intelligence doesn't meet the basic prerequisites for the application of no-fault liability. Although service providers can more efficiently control technical risks, generative artificial intelligence generally doesn't pose a high degree of danger. It is not advisable to require service providers to internalize all risks of infringement out of consideration for protecting others' rights and interests. Given the need to alleviate the burden of proof on victims and moderately strengthen the obligations and responsibilities of service providers, the presumed fault liability should be introduced. Service providers can rebut the presumption of fault by proving that they have fulfilled the duty of care related to avoiding infringing output at both the training and the output stages. The standard of the duty of care should be appropriately strict to incentivize service providers to take adequate risk prevention measures.

Keywords: generative Al ; tort liability ; principles of liability attribution ; presumed-fault liability