The rapid advancement and widespread adoption of AI technologies are embedding intelligence throughout communication systems, particularly in the design of AI-native 6G architectures. This integration of AI into beyond 5G/6G networks and user equipment (UE) presents both opportunities and challenges with respect to security and privacy. These AI models’ potential is both a blessing and a curse, making them a “double-edged sword”. While they can be leveraged to enhance security measures, malicious actors to launch sophisticated attacks can also exploit them. We will discuss in detail about both sides of the coin in this chapter, (a) AI for Security: Proactive Threat Detection and (b) Security for AI: Privacy Preserving Personalization. AI’s role in enhancing security is significant, enabling proactive threat detection, automated incident response, and real-time mitigation. LLMs, such as OpenAI’s GPT models, can analyse vast datasets to identify patterns, anomalies, and emerging threats, complementing traditional cybersecurity tools like firewalls and encryption. This system can be extended to automated incident response, generating security policy commands and network fuzzing test cases for preventive predictive analysis. While AI enhances security, it also introduces vulnerabilities. Security has evolved from predictive AI systems to Generative AI, Agent-based AI, and finally Quantum AI, with each phase introducing new challenges and advancements in safeguarding AI technologies while integrating into future communication systems. We will discuss in details about security threats and measures to be taken for these various stages of evolution of AI security landscape further in this chapter.