The influence of Agentic AI, autonomous systems that act and interact with one another, is growing rapidly as these technologies advance and adoption accelerates. While they enable new forms of automation and intelligence, they also pose risks when agents collude, manipulate, or coordinate covertly. Our research develops methods to detect and defend against such behaviors: from collusive agents in markets that undermine stability, to AI-driven activity in social media that amplifies misinformation. By combining network theory, temporal modeling, and machine learning, we aim to design defenses that safeguard both systems and individuals in this new technological era.