Cyber-Physical Systems (CPS) incorporating Artificial Intelligence (AI) are revolutionizing industries; however, their dependence on connected devices comes with considerable security and safety threats. CPS running in domains such as autonomous vehicles and smart healthcare process large-scale data and demand secure and real-time networking. Traditional security methods are insufficient for CPS, rendering them vulnerable to threats such as data tampering, spoofing, and adversarial attacks on AI models, which could compromise system integrity and user privacy. This chapter tackles the issue of secure communication in AI-based CPS. First, we introduce a security architecture that incorporates state-of-the-art encryption methods, secure communication protocols, and then AI-driven anomaly detection techniques. This chapter also outlines several avenues for future research, including standardizing security protocols and crafting regulatory frameworks to facilitate secure CPS deployment over various domains. Upon addressing these challenges, the chapter provides practical guidelines for researchers, engineers, and policymakers to enhance the safety and dependability of the AI-based solutions integrated into CPS in a data-avant-garde domain.