Agentic AI systems don’t just predict — they act. They plan, reason, call tools, move data, and trigger workflows autonomously. With that autonomy comes a radically expanded attack surface. Based on the OWASP Top 10 for Agentic Applications (2026), this session dives deep into five of the most critical and consistently exploited risks facing modern agentic architectures.
In this technical session, you’ll learn how attackers subvert autonomous behavior, chain multi‑step exploit paths, and weaponize the very capabilities that make agentic systems powerful. Through real‑world examples, architectural breakdowns, and practical mitigation strategies, you’ll leave with a concrete blueprint for securing your own agentic applications.