SOC it to me: how to blueprint AI security governance using the tools you have

AI agents are entering the SOC fast but most teams are not securing them. This session focuses on the real risks of deploying Microsoft Security Copilot and agentic AI in production environments including prompt injection, data exposure, over permissioned connectors, and uncontrolled autonomy. You will walk away with practical blueprints and proven design patterns for secure AI adoption including least privilege architecture, RBAC alignment, auditability, prompt governance, and safe MCP integration strategies. If you are bringing AI into security operations, this talk shows you how to build it securely from day one.