More peers in my network are exploring AI, especially agents, in security workflows. Unlike LLMs, agentic systems add complexity and expand the attack surface. Here’s how I define and analyze agents in my Microsoft research. applied-gai-in-security.ghost.io/agents-clari...