AI agents

AI agents are rapidly moving from labs into production systems, where they can plan, call tools, and act autonomously across critical workflows at machine speed. This talk explores how that autonomy reshapes the attack surface through threats like prompt injection, tool misuse, data exfiltration, and credential theft, and why traditional appsec and IAM controls are not enough on their own. Attendees will learn a practical threat model and a set of concrete defenses—from least‑privilege tool design and MCP/LLM hardening to runtime monitoring of agent behavior—that they can apply immediately to secure emerging agentic architectures.

Presenter: Rod Soto