Governance-Aware Federated Learning for Trustworthy and Compliant Decentralized Infrastructure Monitoring

The paper introduces a federated learning (GFL) architecture that is governed to monitor decentralized infrastructure in a legally heterogeneous setting. The framework incorporates legal compliance limits, auditability controls, and policy alignment as governed by LLM into the federated optimization process, thus creating the possibility of trustful and policy-oriented deployment of AI to the nodes of the public sector.This is to develop a mathematical model that embodies multi-agent lawful infestations, metadata audit rating, and also dynamic trust stabilization within the limits of regulations. Empirical analysis with a synthesized dataset of 20 European jurisdiction nodes reveals that the GFL model performs better than governance-free baselines in the accuracy of their predictions +1 3.5 %) and concurrently enhances transparency, explainability, and auditability. Clients that follow a legal and semantic protocol of governance always provide quality and more consistent updates. The findings show that rather than deterring model convergence and trust in the decentralized systems, the enforcement of governance improves such aspects. The paper is another addition to the intersection of federated AI in digital governance, providing a scalable strategy to institutional compliance in critical areas of infrastructure, such as smart grids, urban mobility, and environmental sensing