Understand why access controls built for people and static workloads fall short for autonomous AI agents.
Learn what to design at the design stage, before an agent is built: its identity and who it acts on behalf of, the tasks and tools it is allowed, the policy it must follow, its risk level, and where a human must approve..
Explore how authorizing each action on intent and behavior stops drift, prompt injection and privilege creep before the action runs.
See how controls can support EU AI Act and GDPR compliance.
Take away a prioritized starting point, from monitor mode to enforcement.
Enterprises are moving autonomous AI agents from pilots into production. These agents call APIs and tools, and hand work to other agents at machine speed. A role defines what an agent may touch, not whether an action still serves its task, so an agent can pass every permission check and still drift, be steered by content it reads, or inherit more access than it needs.
Identity teams and AI security teams each cover part of the problem, and neither covers the whole lifecycle. This webinar presents a practical reference architecture that follows an agent from design to runtime. It covers how authority is delegated and narrowed during hand-off between agents, how credential brokering replaces standing secrets with short-lived, task-scoped credentials, and how authorizing each action on intent and behavior stops drift, prompt injection and privilege creep before the action runs. It ends with a gradual path from monitoring to enforcement that helps produce the evidence the EU AI Act and GDPR call for.
John Tolbert, Principal Analyst and Practice Lead Cyber Security at KuppingerCole Analysts, will give the independent view on how the AI agent lifecycle maps to identity and AI security controls. He will explain why agents should be treated as first-class principals with their own Agent IDs, then walk through the controls at each stage. John will show how intent can inform authorization decisions and identify the common gaps between identity programs and AI security programs.
Yash Prakash, Co-Founder and Chief Strategy Officer at Reva.AI, will show how intent- and behavior-based access control, runtime authorization, and threat detection fit into the architecture.
Magnus Jungsbluth, Senior Principal Engineer at Zalando, will share the practitioner perspective on moving agents into production while balancing speed, security, and compliance.
Who Should Attend
This webinar is designed for CISOs, IAM and security architects, AI and platform security leaders, and compliance professionals responsible for securing AI agents.
See All Locations
See All Locations