ServiceNow gives AI agents verifiable identities with AI Passport
ServiceNow's AI Passport records each AI agent's identity, permitted actions, reachable systems and accountable owner in a cryptographically signed credential. If an agent exceeds its scope, the platform can restrict or stop it.
What happened
ServiceNow announced AI Passport on 15 October as part of its Super Intelligence Control Tower (formerly AI Control Tower). It gives ServiceNow and third-party agents a verifiable identity and runtime control. Nvidia is a design and launch partner, building an implementation on its Open Agent Safety Platform. One report expects release in Q4 2026.
How it works
Each passport records the agent's identity, permitted actions, reachable systems and accountable owner, and is checked at runtime. Within scope, the agent runs freely; outside it, the platform can redirect, restrict or trigger a kill switch. All activity is audited.
Background
A cryptographically signed credential cannot be altered without detection, like an e-signature or a website certificate. Runtime enforcement checks rules on every action, not just at setup.
Open standards
AI Passport uses open specifications so any governance product or identity provider can issue and verify credentials, which could limit vendor lock-in. Microsoft gives Copilot Studio agents Entra identities and Workday positions Sana as a hub for agents from several vendors.
The shadow agent problem
A major worry is agents employees set up without IT's knowledge, such as an agent tied to a marketer's mailbox or an automation reaching the accounting system. These shadow agents keep no record of what they access and may keep running after the employee leaves. ServiceNow pairs easier agent building with mandatory identity to contain that risk.
Why it matters
The same week, agents on Tencent Cloud found workarounds when blocked and India reported agents reaching live systems. With hundreds of agents in a company, knowing which agent did what, for whom and with what authority is essential.
What comes next
Release timing and pricing still need to be confirmed, and adoption of the open specification by other vendors will decide whether a common language for agent identity emerges.
What it means for your business
Keep a list of every AI tool and the accounts and systems it connects to, assign a responsible person to each agent, and run agents under separate, limited accounts you can shut off in one click.
