Embedded Files
Securing AI Agents
Securing AI Agents
By: Mohan Kumar, Co-founder & CEO of Aira Security
By: Mohan Kumar, Co-founder & CEO of Aira Security
Abstract:
Abstract:
AI agents are far riskier than simple chat bots because they take real actions across systems. This talk breaks down the modern agent stack and shows where things fail in practice, including goal tampering, memory poisoning, auth hijacking, and unsafe tool use. A live demo illustrates how these weaknesses can turn a normal user into an admin and leak sensitive data. The takeaway is that teams should give agents only the access they need, enforce delegated authentication, and monitor every action as if it matters.
About the Speaker:
About the Speaker:
Mohan Kumar is the co-founder and CEO of Aira Security focused on runtime protection for AI agents. He previously spent six years at Box working on production security and AI infra. He constantly seeks ways to elevate security processes and culture to the next level.
Chapter Elections & 2025 Year in Review
Chapter Elections & 2025 Year in Review
By: ISC2 SV Board of Directors
By: ISC2 SV Board of Directors
Page updated
Report abuse