This article introduces the Agentic AI Security Scoping Matrix, a framework for securing autonomous AI systems that extends beyond traditional generative AI security approaches.

• Agentic AI systems autonomously execute multi-step tasks, make decisions, and interact with infrastructure and data
• Four scopes defined by agency levels and autonomy: No agency, Prescribed agency, Supervised agency, Full agency
• Agency refers to permitted actions within operating environment; autonomy refers to independent decision-making without human intervention
• Six critical security dimensions: identity context, data/memory/state protection, audit/logging, agent controls, agency perimeters, orchestration
• Security requirements escalate across scopes from workflow integrity to continuous behavioral validation
• Key patterns: progressive autonomy deployment, layered security, continuous validation loops, human oversight integration, graceful degradation

Organizations should assess current agentic use cases, identify capability gaps, develop progressive deployment strategies, and implement continuous monitoring appropriate for their scope level.