From Formal Regulation to Operational Capacity: Asymmetric Adaptive Coordination in Deepfake Governance

Abstract

Deepfake governance is often approached as a problem of technical detection, legal prohibition, or platform responsibility. Yet formal regulation and multi-actor participation do not necessarily translate into effective governance capacity. This article examines why formally dense governance arrangements can remain operationally weak through a qualitative comparison of the Huangshi AI voice-impersonation case and the Cangnan AI face-swapping case. Drawing on publicly available police, judicial and procuratorial materials, regulatory documents, authoritative news reports, and platform or industry sources, the study combines structured-focused comparison with process tracing. The analysis identifies four interconnected mechanisms: temporal mismatch between technological change and institutional updating; jurisdictional fragmentation across the harm chain; regulatory dependence arising from capability and information asymmetries; and incentive divergence that makes collaboration selective and episodic. The findings show that governance weakness arises not simply from the absence of rules or actors, but from failures to convert dispersed information into timely, verifiable, and coordinated action and to translate case responses into institutional learning. The article develops asymmetric adaptive coordination as a mechanism linking technological turbulence, distributed authority, unequal information control, and divergent incentives. It argues that effective deepfake governance depends less on adding rules or participants than on building capacities for timely coordination, auditability, learning, and correction while safeguarding security, privacy, due process, innovation, and public trust.