Securing & Governing AI Agents

​Free for the first 10 seats, then $25. Two hours, live on Zoom. Eligible for 2 self-reported CPEs (ISACA, ISC2).

​Hosted by Meenu Chadha, fractional vCISO and founder of Cyber Advisory. 13 years in cybersecurity GRC at JPMorgan Chase, EY, S&P Ratings, MUFG and Cantor Fitzgerald.

​Okta surveyed 292 executives and 492 workers in March 2026. 92 percent already have AI agents in use. 47 percent have no AI governance strategy at all. 58 percent have already had an AI-related security incident or near miss. Only 34 percent apply the same controls to agents that they apply to humans.

​The deployment is racing ahead. The governance is not. Autonomous agents that read data, call tools and act on your behalf are becoming both a leading source of data leaks and a favorite new attack surface.

​This is not a "what is an agent" talk. It is a working session where you build a real governance and security control set for the agents already in your stack.

​What we will build together:

• ​An agent threat model using the OWASP agentic risks: goal hijacking, tool misuse, over-permissioning, memory poisoning and rogue autonomy

• ​Identity and least-privilege controls so an agent can only touch what it should

• ​Human-in-the-loop gates for high-impact actions, plus logging and monitoring that hold up in an audit

• ​A control set mapped to ISO 42001 and NIST AI RMF, so the work counts toward the frameworks your buyers already ask about

​Who it is for: founders and CTOs shipping agents, security and GRC leads being asked to sign off on them, and anyone running agents in production without a control set.

​You will leave with a working agent-governance control set, plus the option to book a free 30-minute gap check on your own program.