Abstract:

China's regulatory sandbox for artificial intelligence currently prioritizes safety maintenance over fostering innovation, particularly evidenced by the absence of a legal liability exemption system. Regarding institutional provisions, although the inclusive and prudent policies implemented for the AI industry contain implications of liability exemption and relevant legal provisions leave room for such exemptions, the specific regulatory rules lack institutional arrangements for liability exemption. This results in artificial intelligence providers participating in sandbox testing not receiving innovation incentives or preferential treatment through liability exemptions.In response, it is necessary to ground the approach in the theory of objective imputation, clarify the theoretical logic and legal foundation for liability exemptions, and define the scope of exemptions-specifically, excluding exemptions for civil and criminal liabilities while allowing discretionary exemptions for administrative liabilities.Building on this foundation, rigorously define the substantive requirements for liability exemptions and enhance exemption procedures through specifying eligible subjects for exemptions,designing binding guidance mechanisms and establishing oversight and relief measures to establish a legal liability exemption system tailored to the practical and developmental needs of China's AI regulatory sandbox.

Key words: artificial intelligence, regulatory sandbox, exemption from responsibility, administrative responsibility