Autonomous AI agents can plan, use tools, access data and alter production systems with limited human approval. Yet many security and governance programs were designed for static AI systems and lack the runtime controls needed to govern independent action. As agent adoption accelerates, organizations need a security architecture that addresses the full agent stack while establishing clear accountability.

Led by Sunil Arora of Humana, this session will cover:

• How the CIAA model extends the traditional security triad - confidentiality, integrity and availability - by making accountability a foundational requirement;
• How the seven-layer MICE framework applies defense in depth across infrastructure, data, models, agent execution, identity, monitoring and governance;
• What CISOs, security engineers and governance boards should prioritize during the first 90 days of securing production agent workflows.