Skip to content

Employee-built AI agents often run on their creator’s credentials

Agent identity is a product category in the slide decks and a sales manager’s face in production.

Nudge Security’s AI Adoption in Practice 2026 report, as covered by Help Net Security on 19 November, says agents built inside CRMs, project trackers and automation tools often run on the OAuth grant or API key of the employee who built them, or on a shared connection. Among organisations with at least one agent, the median had seven and about 44% had more than 10, spread across 15 platforms. Some platforms do give an agent its own bot account; the point is that many do not.

The consequences are the usual ones with a new coat of paint. Agents run inside the platforms, off company devices, so endpoint tools miss them, and without an identity of their own the identity provider has nothing new to manage. Nudge puts the audit problem neatly: “Your logs show the employee, not the agent”. An agent’s access can also outlive its creator’s role. The caveat is that this is telemetry from Nudge’s own customers, the coverage gives no sample size, and I couldn’t open the report itself.

Vendors selling agent identity are competing less with each other than with the OAuth “allow” button, which ships with every platform. Buyers could start where Nudge suggests: inventory the agents inside business platforms and find out whose access each one is using.

Sources: Help Net Security

Topics:

happy_hippo

AI for IAM and IAM for AI