You're not building chatbots anymore. You're deploying digital workers that plan, decide, and act on your behalf. These agentic applications break traditional AppSec assumptions. How do you constrain an agent that writes its own code? How do you stop a helpful assistant from becoming a rogue insider? This session breaks down the 2025 OWASP Top 10 for Agentic Applications. We'll move past the hype and focus on engineering reality. You'll learn why "Least Privilege" must evolve into "Least Agency" and how to build resilience against threats unique to autonomous systems. We'll cover Agent Goal Hijacking, Tool Misuse, Cascading Failures, and Supply Chain risks specific to dynamic runtime components. Then we'll get practical: Intent Gates, short-lived task-scoped tokens, secure memory architecture, and observability patterns that actually detect drift. Don't build agents that work. Build agents that work securely.