Abstract：Ethical review of artificial intelligence technology is a key institutional arrangement for achieving effective governance of artificial intelligence.The current review practice faces three core dilemmas.The review standards are static and rigid,resulting in insufficient dynamic adaptability,making it difficult to cope with the rapid iteration of technology and complex application scenarios.Due to the lack of interdisciplinary knowledge supply and delayed updates during the review process,risk assessment deviates from diverse value cognition and social consensus.The review mechanism has fallen into a fragmented governance situation with mixed functions and weakened effectiveness due to the failure to connect with other risk governance tools.As a dynamic governance paradigm that quickly adapts to environmental changes,agile governance has core advantages such as dynamism,adaptability,and synergy,which meet the governance needs of ethical risks in generative artificial intelligence.It can optimize the institutional review of artificial intelligence ethics through the construction of a dynamic typification standard system,the embedding of ethical negotiation mechanisms,and the application of a regulatory model that combines software and hardware.Guided by this,the ethical review system for artificial intelligence should establish a dynamic standard system that is agile,iterative,and scenario oriented,as well as an open,collaborative,and continuously evolving knowledge supply mechanism.Based on this,it should achieve effective connection between ethical review and other risk governance mechanisms,and promote the efficiency of ethical risk governance for artificial intelligence.
孙方磊. 迈向敏捷治理:人工智能伦理审查的实践困境与体系优化[J]. 中国科技论坛, 2025(11): 130-139.
Sun Fanglei. Moving Towards Agile Governance:Practical Challenges and System Optimization in Ethical Review of Artificial Intelligence. , 2025(11): 130-139.
2025, 
