Achieving accurate and secure data dissemination in dynamic VANETs requires expressive and conflict-free access control policies. To address this, we propose a policy enforcement framework that enables high-mobility vehicles and RSUs to co-design access control policies using disjunctive normal form (DNF) for flexibility and accuracy. Conflicts are resolved through confidence-weight-based mechanisms. Additionally, we introduce RLID-V, a reinforcement learning-based policy generation scheme that dynamically updates confidence weights and incorporates decision tree-based feedback for continuous improvement. Experiments in traffic guidance and accident warning scenarios demonstrate enhanced accuracy, robustness, and negligible delay overhead compared to existing methods, making this approach a reliable solution for secure data dissemination in VANETs.