The increasing risk of data breaches necessitates advanced security solutions, with AI-driven Data Leakage Prevention (DLP) systems emerging as a key defense mechanism. This framework integrates deep learning techniques, including CNNs, LSTMs, autoencoders, and GANs, for anomaly detection and simulated attacks. Threat detection is enhanced through anomaly scoring, incident response, and policy enforcement. Additionally, security is fortified with federated learning, blockchain for audit and integrity, and homomorphic encryption for privacy preservation. This hybrid approach ensures scalable, interpretable, and resilient data protection against evolving cyber threats. Future research should focus on optimizing computational efficiency while maintaining high detection accuracy and privacy assurance