The EU Artificial Intelligence Act establishes an artificial intelligence regulatory sandbox through unified legislation, promoting its implementation across the European Union. As an important practice of experimental legislation, the EU AI regulatory sandbox integrates the concepts of dynamic regulation and the precautionary principle. It demonstrates institutional advantages in areas such as objective setting, participation mechanisms, operational coordination, and data processing rules, while expanding the scope of AI testing through coordination with the General Data Protection Regulation (GDPR). Nevertheless, the EU AI regulatory sandbox still faces challenges concerning liability exemption mechanisms and the prevention of regulatory rent-seeking.In comparison, although China’s artificial intelligence industry has experienced rapid development, its AI regulatory sandbox mechanism and related practices of experimental legislation remain at an exploratory stage. By examining the institutional design, experimental legislative characteristics, and limitations of the EU AI regulatory sandbox, this paper provides insights for the future development of China’s AI regulatory sandbox framework.