Generative AI (GenAI) is reshaping the nature of cyber threats, particularly against military cyber-physical systems (CPS) such as drones and radar networks. These AI-driven attacks generate highly deceptive data, allowing hostile entities to appear friendly or go undetected—posing critical risks in real-time defense operations. Traditional detection methods that rely on pattern anomalies or statistical shifts often fail against such sophisticated threats. This research presents a novel approach that uses GenAI not only to simulate potential cyberattacks but also to enhance system defenses. A detection framework was developed that leverages GenAI to create diverse and realistic attack data, enabling defense models to learn and adapt more effectively. The system achieved an 85% detection rate, outperforming conventional techniques. For prevention, GenAI-generated attack scenarios were used to train defense AIs in real time, reducing successful intrusions by 70%. Additionally, a GenAI-based recovery module was implemented to correct compromised data—such as distorted radar signals—restoring system accuracy to 90% within one second of attack. Experiments in a simulated military environment showed that attacks which previously succeeded 85% of the time were significantly neutralized. While further improvements are needed to address complex attack vectors and optimize processing speed, the findings demonstrate that GenAI can serve as a powerful ally in cyber defense. This study offers a strategic framework for securing military CPS, using AI’s own capabilities to outmaneuver evolving cyber threats and enhance operational resilience.