The OWASP Top 10 for Agentic AI Applications identifies what goes wrong when autonomous agents are attacked. This session examines where it goes wrong and why that distinction determines whether your governance controls can reach it. Drawing on the full OWASP Agentic AI Top 10, this session establishes the reasoning layer as the common attack surface across ASI01 through ASI09, demonstrates why execution-layer controls cannot intercept reasoning-layer compromise, and gives practitioners three governance moves they can implement immediately.