Learn to secure autonomous AI systems and multi-agent architectures through this hands-on conference talk from DEF CON 33. Explore the expanded attack surface of agentic AI systems, including vulnerabilities in tool execution, persistent memory, and inter-agent communication. Master threat modeling techniques specifically designed for multi-agent pipelines and implement robust supply-chain defenses. Discover methods to safeguard sensitive workflows while addressing prompt injection attacks at scale. Examine real-world case studies featuring both independent and integrated assistant deployments to understand practical security challenges. Implement policy-as-code guardrails and fine-grained access controls to protect AI systems without hindering innovation. Develop red-team strategies for testing agent behavior and learn actionable patterns for defending end-to-end autonomous systems. Gain comprehensive knowledge applicable whether you're securing existing AI workflows or conducting penetration testing on agentic systems.