Security by design for agentic AI: How CISOs stay in control of autonomous systems

Agentic AI shifts the security problem from protecting systems to governing autonomous behavior. CISOs and platform engineering leaders are already seeing this change as AI agents deploy infrastructure, modify configurations, and make operational decisions at machine speed. Traditional controls break down not because they are wrong, but because they were built for humans, not software acting independently.This talk draws on emerging enterprise patterns where security is embedded directly into internal developer platforms to manage agentic risk. We will explore how leading organizations are redefining the relationship between CISOs and platform teams, moving from policy enforcement to guardrail design, from point-in-time approvals to continuous risk signaling, and from dashboards to decision-oriented narratives.The session also introduces the concept of an AI Chief of Staff for the CISO: a way to translate platform-level signals into clear priorities, tradeoffs, and accountability. The goal is not more control, but the ability to operate with confidence in systems where autonomy is unavoidable.