From Algorithmic Detection to Adaptive Fisheries Governance: An AI-Enabled Compliance and Ecosystem Feedback Framework for IUU Fishing in the Yellow Sea

Artificial intelligence (AI) is increasingly able to infer apparent fishing activity, vessel type, encounters, non-cooperative or otherwise low-visibility vessel presence, and other maritime patterns from Automatic Identification System (AIS), vessel monitoring system (VMS), radar, optical imagery and catch records. Yet better detection does not automatically produce better fisheries management. This article asks under what institutional conditions AI-enabled monitoring can support more effective compliance and the sustainability of shared fish stocks without converting probabilistic signals into automatic enforcement findings. Using the Korea–China Yellow Sea as an institutional case study, the analysis combines recent AI-fisheries literature, international monitoring–control–surveillance guidance, ecosystem-based fisheries scholarship, the Korea–China Fisheries Agreement and verifiable official enforcement and resource-management developments through August 2026. It develops an Adaptive Fisheries Intelligence Loop (AFIL) with five stages: observe, validate, respond, learn and adapt. The model requires human validation of algorithmic signals, provenance labels for shared information, feedback from enforcement outcomes, and a second feedback channel connecting compliance patterns to ecological indicators and management measures. Applied to the Yellow Sea, AFIL supports joint risk vocabularies, interoperable case and vessel histories, targeted patrols, scientific stock assessment and periodic recalibration of both algorithms and management rules. AFIL is an institutionally plausible and testable governance design, not an empirically validated prediction that AI will itself deliver compliance. The article therefore argues that AI should be governed as adaptive decision-support infrastructure rather than as an automated enforcement endpoint.