AI Agent Permissions Need an Accountable Owner

On October 15, 2026, ServiceNow announced AI Passport, a system intended to verify what AI agents are authorized to do during workflows, alongside expanded discovery and governance tools. The company says AI Passport and some other capabilities are expected in Q4 2026—not that all have already launched. Read the announcement.

The useful shift is from recording what an agent did to setting and checking its authorized scope as it acts. That could make responsibility clearer. But a permission system cannot decide by itself who should grant authority, whether the boundaries fit the work, or who answers when they don’t. Those choices still need named owners and meaningful review.

A practical test for any such rollout: can the responsible person explain the agent’s allowed actions, the limits, and how to pause or revoke access?