Relational Database Management Systems (RDBMS) are foundational for structured data management across various domains, making their security paramount, especially in the era of Artificial Intelligence (AI). This paper proposes an AI-Driven Access Control (AI-DAC) framework that integrates machine learning (ML), reinforcement learning (RL), and metaheuristic algorithms to enhance data access control in RDBMS. The framework is designed to dynamically adapt to evolving security threats, providing a robust, multi-layered approach that includes diagnostic, corrective, and generative loops. The AI-DAC framework's effectiveness is validated through comprehensive experimental studies and real-world case analyses, demonstrating its capability to optimize access control policies, reduce unauthorized access, and ensure high security in RDBMS environments.