Abstract:
In the debate over whether China should enact a unified artificial intelligence law, existing scholarship often subsumes questions operating at different levels under the single concept of “unified legislation”. Yet “unification” encompasses at least four dimensions: the hierarchy of legal norms, the breadth of regulated content, the regulatory granularity, and legislative phasing. The hierarchy of legal norms can be reframed as a question of the allocation of legislative authority and normative effect; the breadth of content as a question of legislative model, particularly the relationship between comprehensive and sectoral legislation; and the regulatory granularity as a question of legislative strategy concerning regulatory intensity and the degree of rule specification. Legislative phasing, however, is not absorbed by the first three dimensions. Rather, it constitutes a dynamic axis running through the allocation of legislative authority, the selection of legislative models, and adjustments in regulatory density, and it directly addresses when artificial intelligence legislation should begin and how it should proceed in stages. It is precisely in relation to legislative phasing that artificial intelligence legislation confronts the classic Collingridge dilemma. Accordingly, AI legislation should not be conceived as a one-off decision made at a single point in time, but as a layered, incremental, and feedback-responsive institutional process. Such a process would allow the hierarchy of norms, the legislative model, and the regulatory granularity to be continuously adjusted in light of technological maturity and changing risk structures.
DOI: 10.35534/al.0803009 (registering DOI)
Cite: Jiong, Y. (2026). The Collingridge Dilemma in Artificial Intelligence Legislation and the Incremental Response. Advance in Law,8(3), 113-126.