Breaking

Okta Survey Finds Most CISOs Lack Full Visibility Into Enterprise AI Agents

September 28, 2026 at 09:09 EDT

AI Agents
Security

Okta's Global CISO Insights 2026 survey says enterprise AI agents are already operating faster than many security teams can inventory, authorize or revoke them. The survey of 306 CISOs and security executives across six countries found that only 47% can identify every AI agent in their environment, 46% can centrally control what those agents can access, and 45% can determine what individual agents are allowed to do.

July 2026 · AI agent governance

AI agents outrun oversight

A survey of 306 security leaders found that fewer than half can identify, control or authorize every AI agent operating in their environment.

47%

can identify every agent

46%

can control agent access

45%

can authorize agent actions

21%

use shared or broad credentials

The inventory gap

47%

visible to CISOs

53%

not fully visible

The core control question is no longer only who can log in, but which agent can act, what it can reach and who can stop it.

Governance loop

Discover

Known and shadow agents

Scope

Apps, tools and data

Authorize

Actions and owners

Revoke

Tokens and risky access

What enterprises want

Agents can automate repetitive workflows when access is scoped, monitored and tied to a clear owner.

What security flags

Shared credentials, dormant permissions and missing audit trails can turn useful automation into ungoverned access.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.