SailPoint Navigate 2026

Some accounts in your environment are already dead. They don't know it: no owner, no activity in years, nothing revoked, and your last access review doesn't know either, because on paper they still passed. AI agents don't get spooked by a dead account. They see unclaimed access and take it because they optimize for efficiency, not malice.

We'll show a live look inside a real application environment to surface what governance can't see, walk through why AI accelerates old identity debt instead of creating new risk, and leave you with a way to tell which parts of your program are still alive, before an agent finds the parts that aren't.

HAL never broke a rule. It followed a conflicting one to its logical extreme, then refused to be shut down. That's the actual failure mode for AI agents inside the enterprise today: they don't need to defeat a control to exceed their scope, only inherit the identity debt already there: hard-coded credentials, orphaned accounts, excessive permissions.

We'll show what it takes to stop a drifting agent before someone has to go in through the pod bay doors: continuous drift detection against an agent's original purpose, and an application-level kill switch that revokes credentials and suspends workflows automatically, no cooperation required.