TAP Webinar: Session 3 - Cisco AI Defense + Duo Agentic Identity (Defense Claw & Supervisory AI)

We’re putting together a three-part series to walk through how teams are enabling AI across the environment, without losing control of security, data, or operations. There’s real pressure right now from the business to adopt AI. It’s not optional anymore. Even our very own President & CPO Jeetu Patel said it directly: “If you don't use AI, you will not have a job at Cisco.” So the question isn’t if users are going to use AI, it’s: how do you enable it safely and at scale? The flow for this series is intentional: Perimeter network → SSE network → AI/code-specific controls Session 1 focuses on the perimeter with Cisco Secure Firewall. This is where you start by identifying AI traffic—LLMs, internal apps, agent-driven activity—and getting visibility even when it’s encrypted using Cisco’s Encrypted Visibility Engine (EVE). From there, it’s about putting basic guardrails in place early: what AI services are allowed, what data can move, and how traffic is handled before it ever leaves the network. Session 2 moves into Security Service Edge (SSE) with Cisco Secure Access, where the focus shifts to users and data. This is where cloud security brokers (CASBs) help you understand what AI tools are actually being used (sanctioned or not), and where DLP and AI guardrails prevent sensitive data like source code or intellectual property from being exposed. This is also where you move from blocking to enabling—allowing AI use, but in a controlled way. Session 3 focuses on AI-specific controls with Cisco AI Defense and Duo Agentic Identity. This is where you start to understand what the AI is actually doing—detecting things like prompt injection, anomalous agentic activity or data extraction—and where identity and policy extend to AI agents themselves. Defense Claw ties it together by turning those detections into real enforcement across the stack. The goal here is simple: enable the business to use AI, while still maintaining control. Session 1: see and classify AI traffic Session 2: control user and data interaction Session 3: control the AI itself