Why are AI agents considered a security risk in enterprises?

AI agents operate with access to core systems and sensitive data, sometimes reading, writing, and making decisions, but most are unmanaged and lack the governance controls applied to human identities. This creates visibility and governance gaps that expose enterprises to risk.

What is the first step organizations should take to manage AI agents securely?

The first step is discovery—understanding what AI agents exist in the enterprise, who created them, when they were created, what access they have, and what systems they can touch. Without this visibility, effective governance is impossible.

Managing AI Agents as Non-Human Identities

Name: Managing AI Agents as Non-Human Identities
Uploaded: 2026-04-25T19:13:01-04:00
Duration: 1 min 55 s
Description: TL;DR AI agents are proliferating in enterprises and require access to systems and data like human workers, but most operate without governance controls applied to human identities. The core problem is visibility—AI agents are created and forgotten, op...

pathlock

04/25/2026

0 (0%)

Report Like Favorite

TL;DR

AI agents are proliferating in enterprises and require access to systems and data like human workers, but most operate without governance controls applied to human identities.
The core problem is visibility—AI agents are created and forgotten, operating with access to sensitive data and making decisions without proper oversight or management.
Discovery is the foundational first step: enterprises must identify all AI agents, understand who created them, when, and what access they have before implementing governance controls.

Summary

This executive insight addresses the emerging challenge of AI agent proliferation in enterprise environments. Haviv Rosh explains that AI agents function like human workers by requiring access to applications, systems, and sensitive data to perform their tasks. The core problem enterprises face is that most AI agents remain unmanaged—created by humans but operating with access to critical systems and data without the governance controls applied to human identities. This creates significant visibility and governance gaps. Rosh emphasizes that the first step in securing AI agents is discovery: understanding how many agents exist, who created them, when they were deployed, and what access privileges they hold. Without this foundational visibility, organizations cannot implement effective governance controls. The challenge extends beyond AI agents to include service accounts, bots, and other non-human identities that often proliferate unchecked across enterprise systems.

Chapters

0:00 - AI Agent Explosion
0:23 - The Unmanaged Agent Problem
1:04 - Discovery as Foundation

Key Quotes

0:28 "The problem is that most AI agents today are unmanaged. They are created by humans, but then they sit there operating with access to core systems, have access to your sensitive data."
0:50 "The real issue isn't AI as a concept, it's that these agents behave like identities, but enterprises aren't treating them like identities yet."
1:42 "You cannot manage what you cannot see."

Categories:

» AI & Machine Learning
» Webinar Library
» Webinar Library » Pathlock
» Cybersecurity » Identity & Access Management (IAM)
» Data Protection

Tags:

Show more Show less

TL;DR

Summary

Chapters

Key Quotes

Action1: Vulnerability Digest--Patch Tuesday & Other Updates

Understanding the True Costs of DIY Data Classification vs. Buying Solutions

Action1: The Remediation Gap: Vulnerability Management in the Age of AI

The AI-Powered VMware Alternative

LATAM: Accelerating Insights on AI Through an Engaging Webinar Series