Securing Agentic AI Systems and Multi-Agent Workflows
Presents threat modeling and defensive guardrails for autonomous multi-agent AI workflows, covering persistent memory, tool execution, and prompt injection.
Summary
The Shift from AI Copilots to Autonomous Agents
Artificial intelligence architectures are rapidly transitioning from single-turn copilots to autonomous multi-agent systems that coordinate complex actions across persistent memories, API tools, and downstream databases.
Expanded Attack Surfaces in Multi-Agent Workflows
Andra Lezza and Jeremiah Edwards analyze novel attack vectors in agentic ecosystems, including indirect prompt injection at scale, malicious tool invocations, and compromised inter-agent messaging pipelines.
Implementing Policy-as-Code Guardrails
Securing autonomous workflows requires deploying policy-as-code guardrails, strict least-privilege API permissions, and continuous behavioral anomaly detection across agent execution chains.
Red Teaming and Defensive Blueprints
Case studies demonstrate practical red-teaming methodologies for autonomous assistants and provide defensive blueprints to harden agentic pipelines without hindering developer innovation.
Related talks
Main Stage·46 min
Ben Nassi · Or Yair · Stav Cohen
Demonstrates an indirect prompt injection attack where a Google Calendar invite autonomously triggers Gemini AI agents to exfiltrate private user emails and files.
Main Stage·19 min
Tobias Diehl
Demonstrates how adversaries exploit data voids in LLM training corpora and search indexes to manipulate Microsoft Copilot answers and inject untrusted content.
Main Stage·45 min
Bruce Schneier
Bruce Schneier examines how artificial intelligence systems hack human institutions, tax codes, financial markets, and legal frameworks.