Identity Governance And The Challenge Of AI Agents Moving At Machine Speed
Reported by Forbes1 source ↓
One Identity CEO Praerit Garg argues that established identity governance, privileged access management and zero-trust controls can govern AI agents without a separate control architecture. He says agents should not have more permissions than the users they act for.
Key points
- Garg says agents should be limited to the permissions of the users they represent
- Garg cites McKinsey reporting that up to 40% of organizations are scaling an agentic AI system in at least one function
- He says each agent needs its own identity, supervisor and role-specific permissions
- Garg argues that specialized agent controls can add complexity without creating a stronger security boundary
Details
In a Forbes article, One Identity CEO Praerit Garg argues that organizations can use existing identity governance and administration (IGA) and privileged access management (PAM) tools to control AI agents. He says the same access questions that apply to employees should apply to agents, including what data they can reach, which actions they can take and when approval is required.
Garg says an agent acting for an employee should have no more authority than that employee. He argues that each agent should have a distinct identity, a defined supervisor and permissions tied to its role, so organizations can track actions and revoke access without disrupting other agents. He also says privileges should be reviewed and audited because agents can exercise them repeatedly and at high speed.
To illustrate the limits of specialized controls, Garg points to Microsoft’s earlier code access security approach, which the company later deprecated and no longer treats as a security boundary. He says Microsoft instead directs developers to enforceable operating-system boundaries, including user accounts. Garg also cites McKinsey reporting that up to 40% of organizations are scaling an agentic AI system in at least one business function.
Written from coverage as of
Images
- Forbes
Stocks in this story
Change since open: 20 Nov to 21 Nov 2026
Last refresh