Abstract：Systematic ethical review is central to the framework of AI governance.Current practice are characterized by formal compliance oriented toward complete documentation and procedural thoroughness.This manifests in four pathologies.First,ethical review norms suffer from insufficient superior legal basis and fragmented rules,which fundamentally constrains review effectiveness.Second,review bodies are organizationally dependent,making it difficult to form independent judgments consistent with public interests.Third,procedural operation tends to be templated and limited to one-time approval,while follow-up review is perfunctory.Fourth,inadequate responsibility allocation and incentive-restraint mechanisms dampen the motivation for substantive review.Against the tension between formal compliance and substantive governance,this paper proposes a systematic improvement path based on the theoretical pillars of normative legitimacy,structural rationality,and result verifiability.It establishes a comprehensive legislative framework under a“general-specific”structure;builds an independent and hierarchical organizational system with the introduction of third-party review;promotes review standardization using technical verification tools such as documented interpretation and counterfactual reasoning;and establishes a safeguard mechanism that balances responsibility allocation and positive incentives.Ultimately,this will shift ethical review from symbolic compliance to substantive governance centered on risk identification and responsibility implementation.
王保民, 张晨晓. 人工智能伦理审查的体系化完善:从形式合规到实质治理[J]. 中国科技论坛, 2026(5): 142-150.
Wang Baomin, Zhang Chenxiao. Systematic Improvement of AI Ethics Review: From Formal Compliance to Substantive Governance. , 2026(5): 142-150.
2026, 
