AI agents are starting to do more than answer questions. They can search systems, choose tools, update records, trigger workflows, and make changes in business applications.
Once an agent can take action, it should be treated like a system identity. It needs its own identity, a clear owner, limited permissions, and rules that control what it can do.
This talk uses a simple employee-badge analogy to explain the idea. An agent is like an employee or contractor, and its identity is the badge. Least privilege means the badge opens only the doors needed for the job. Bounded access means it works only in the right building, on the right floor, for the right task, and for a limited time.
The session also explains why every action should be checked at runtime, why the target system must still protect its own resources, why access should be short-lived, and why every agent action needs a clear audit trail.
Attendees will leave with a practical way to think about AI agent security: treat the agent as an identity, limit its authority, and check every action before it reaches an enterprise system.
Speaker(s): Venkata Patelkhana
Agenda:
6:25 to 6:35 PM – Open for participants to enter and network.
6:35 to 6:40 PM – IEEE LM and CTCN Business meeting and to introduce speaker.
6:40 to 7:55PM – Formal Program and Q&A.
Virtual: https://events.vtools.ieee.org/m/571370