The increasing digitization of political communication has significantly transformed the scale, speed, andcomplexity of electoral influence operations across global digital ecosystems. State-sponsored disinformationnetworks, coordinated propaganda campaigns, bot-driven amplification systems, and synthetic mediamanipulation increasingly exploit online platforms to shape public perception, destabilize democratic institutions,and influence electoral outcomes across national boundaries. Conventional counter-disinformation mechanismsfrequently rely on centralized moderation architectures and static machine learning models that often struggle toadapt to rapidly evolving multilingual narratives, decentralized propagation behaviors, and adversarialmanipulation tactics. Simultaneously, concerns regarding data sovereignty, privacy protection, and geopoliticaljurisdictional constraints have limited large-scale collaborative intelligence sharing between digital platforms andnational regulatory environments. This study proposes a Federated Deep Reinforcement Learning (FDRL)framework for adaptive counter-disinformation messaging during cross-border electoral influence operations. Theframework integrates federated learning architectures, reinforcement-based policy optimization, multilingualsemantic intelligence, and adaptive communication intervention mechanisms to support decentralized andprivacy-preserving detection and response capabilities across interconnected digital ecosystems. The proposedsystem further incorporates behavioral synchronization analytics, narrative propagation modeling, and adaptiveresponse optimization to counter evolving disinformation strategies in real time. Findings demonstrate thatfederated deep reinforcement learning significantly improves adaptive response efficiency, strengthens resilienceagainst coordinated electoral manipulation campaigns, and enhances scalable counter-disinformation governancewithin globally distributed communication environments.