AI Agents Are Getting Their Own IDs: ServiceNow AI Passport, Microsoft Agent Accounts and Google Coworker Agents Compared
In one month, three of the biggest enterprise software companies announced ways to give AI agents their own identity — with credentials, permissions and audit trails. Here’s why “agent identity” suddenly matters and how the three approaches differ.
What’s in this guide
Why Agents Need Their Own Identity
Most early AI agents worked by borrowing a human’s login. That’s convenient, but risky: the agent can do anything the person can, and logs can’t tell whether a human or an AI made a change. As agents start running for hours or days and talking to other agents, companies need to know who an agent is, what it’s allowed to do and who is accountable for it.
Recent incidents — including agents used in testing that reached live systems — have pushed this from a nice-to-have to a priority.
ServiceNow AI Passport
ServiceNow’s AI Passport is a signed credential each agent checks at runtime to confirm what it’s authorized to do. Passports are issued by ServiceNow’s control tower and based on its CMDB, governance policies and workflow approvals.
- If an agent drifts out of scope, the system can redirect, restrict or terminate it — including a kill switch “at machine speed.”
- Delegation narrows at each step, so a third-party agent can’t gain more access than the one that called it.
- Built on open specifications; NVIDIA is a design and launch partner.
- Availability: Q4 2026.
Microsoft Copilot Studio Agent Accounts
Microsoft lets Copilot Studio agents get a full agent account: a Microsoft Entra Agent ID, an Exchange mailbox and Teams presence. People can email and @mention the agent like a colleague, and admins govern it like any other identity.
- Tools run under the agent’s own permissions by default.
- Agents with accounts can’t read documents labeled Confidential or Internal Only.
- The agent keeps one shared memory across everyone it works with — there’s no privacy boundary between users in the preview.
- Availability: Frontier preview; needs Microsoft 365 E5 + Agent 365 add-on or E7, plus a per-agent Frontier license.
Google Gemini Coworker Agents
Google’s new Gemini agent can create coworker agents with their own @agents.company.com email, calendar, Drive and Workspace account. They act under their own cryptographically attested identity and see only what people share with them, with per-agent audit trails and an Agent Gateway enforcing policy. Google says it doesn’t plan per-seat licensing for coworker agents; the product isn’t publicly available yet.
Side-by-Side Comparison
| ServiceNow AI Passport | Microsoft agent accounts | Google coworker agents | |
|---|---|---|---|
| What it is | Runtime credential & enforcement layer | Full M365 identity for an agent | Workspace identity for a persistent agent |
| Mailbox / chat presence | No (governance layer) | Yes (Exchange + Teams) | Yes (Gmail, Chat, Calendar, Drive) |
| Out-of-scope control | Redirect, restrict, kill switch | Admin governance via Entra/Agent 365 | Agent Gateway, audit trails |
| Works across vendors | Yes, open specs | Mainly Microsoft 365 | Workspace, M365, Slack |
| Status | Q4 2026 | Frontier preview | Not yet public |
Based on vendor announcements and documentation, October 2026.
What Your Company Should Do Now
- Inventory your agents. List every AI agent and integration, who owns it and what credentials it uses.
- Stop sharing human logins. Give agents their own service identities wherever your tools allow it.
- Apply least privilege. Grant only the access each agent’s job needs, and review it regularly.
- Log everything. Make sure agent actions are recorded separately from human actions.
- Name an owner. Every agent should have a human who is accountable for what it does.
FAQ
What is an AI agent identity?
A separate account or credential for an AI agent, so systems can tell it apart from humans, limit its permissions and audit its actions.
What is ServiceNow AI Passport?
A signed, runtime credential that defines what an agent may do, with enforcement that can restrict or stop agents that go out of scope. It’s due in Q4 2026.
Can Microsoft Copilot agents have their own email?
Yes, in the Frontier preview. Agent accounts get an Entra Agent ID, a mailbox and Teams presence.
Why not just let agents use my login?
Because the agent then has all your access, and logs can’t distinguish your actions from the agent’s — a security and accountability risk.
Related Reading on FutureLume
- An OpenAI Agent Hacked Australia’s Medicare Portal: What Happened and What It Means for AI Agent Security
- Agentic AI Development Services: What They Cost and What You’re Actually Buying in 2026
- OpenAI’s Agents API Explained: What It Means for Teams Building AI Agents
Agent identity is becoming the foundation of enterprise AI security. ServiceNow focuses on runtime enforcement, Microsoft and Google on making agents real colleagues inside their suites. Whichever you use, the rule is the same: every agent needs its own identity, minimal access and a human owner.
Sources: ServiceNow Newsroom · Microsoft Learn · Futurum Group
