As artificial intelligence (AI) technologies rapidly penetrate key sectors such as healthcare, education, and public governance, traditional static, ex-post regulatory models can no longer keep pace with the risks and uncertainties arising from their swift evolution. Focusing on the AI regulatory sandbox, the study clarifies its institutional logic, evaluates implementation outcomes, and identifies practical challenges, while exploring how the mechanism can balance innovation with the protection of the public interest. Employing a mixed approach of literature review and comparative case analysis, the study indicates that AI regulatory sandboxes provide a legally bounded, real-world environment that enables iterative testing, risk assessment, and rule refinement, thereby accelerating technological innovation and enhancing regulatory agility, and multi-stakeholder collaboration within the sandbox further strengthens the protection of user rights and social safety. Nonetheless, the mechanism remains challenged by major issues, including regulatory capture, inadequate transparency, fragmented cross-border coordination, and limited regulatory resources. As such, it recommends strengthening information disclosure and legal liability frameworks, establishing mechanisms for international mutual recognition and collaborative governance, and directing targeted investment toward regulatory talent and technical infrastructure.
Zhichao Wang (Wed,) studied this question.