ServiceNow issues each AI agent a signed credential that defines what it may do at runtime
ServiceNow announced AI Passport, a cryptographically signed credential that sets what each AI agent and MCP server may do at every step of a workflow. The prevention claims are ServiceNow's own description of a design.
The Watch · Security desk
Five actors: the Control Tower issues each AI agent its own passport; NVIDIA builds an enforcement implementation; governance products can issue and verify credentials; third-party agents get a verifiable permissions record; drifting agents face redirection, restriction or a kill switch.
- What's now possible capability Control Tower Issues each AI agent its own AI Passport, grounded in the CMDB, GRC policies and workflow approvals, claim 5
- What's now possible capability NVIDIA Design and launch partner building a production-ready enforcement implementation, claim 10
- What's now possible capability Governance products Open specifications let them issue and verify credentials for their own and third-party AI agents, claim 9
- What it limits constraint Third-party AI agents Get a verifiable permissions record rather than relying on the agent's own claims, ServiceNow says, claim 6
- What it limits constraint Agents that drift Can be redirected, restricted or reconnected, up to a kill switch at machine speed, ServiceNow says, claim 4
| Who | How | Kind | Claim |
|---|---|---|---|
| Control Tower | Issues each AI agent its own AI Passport, grounded in the CMDB, GRC policies and workflow approvals | capability | 5 |
| NVIDIA | Design and launch partner building a production-ready enforcement implementation | capability | 10 |
| Governance products | Open specifications let them issue and verify credentials for their own and third-party AI agents | capability | 9 |
| Third-party AI agents | Get a verifiable permissions record rather than relying on the agent's own claims, ServiceNow says | constraint | 6 |
| Agents that drift | Can be redirected, restricted or reconnected, up to a kill switch at machine speed, ServiceNow says | constraint | 4 |
What happened
- If an agent drifts outside its passport scope, ServiceNow says responses range from redirecting or restricting it to a kill switch at machine speed.
- ServiceNow says the passport supports precision revocation and delegation that narrows at each step, with audit trails that carry full compliance signatures.
- NVIDIA is a design and launch partner, building a production-ready enforcement implementation on its Open Agent Safety Platform, OpenShell and Sentry.
Why it matters
- Who has to decide decision Teams constraining production agents have to pick where the check runs: in the agent runtime, at the MCP server, or at the infrastructure layer NVIDIA points to.
- What it limits constraint An agent's scope is only as accurate as the CMDB records, GRC policies and approvals it is built from, so stale records produce stale permissions.
- What's now possible capability If revocation works as described, a team can pull one agent's authority mid-workflow without stopping the other agents in the chain.
The design has three parts. ServiceNow's Super Intelligence Control Tower issues each agent its own passport, built from the CMDB, GRC policies and workflow approvals [5]. The agent consults that credential at runtime, including in environments such as NVIDIA OpenShell, to verify its authorized scope [2]. Enforcement is the third part. ServiceNow sets the whole thing against audit trails, which it says show what happened after the fact [13].
NVIDIA's account of enforcement is specific about location. "As enterprises shift from experimenting with agents to running them in production, securing those agents demands enforcement at the infrastructure layer," said Itay Ozery, director of Product Marketing for Networking and Cybersecurity at NVIDIA [12]. ServiceNow adds that credentials are verifiable at every step of a multi-agent workflow [8].
The layer covers MCP servers as well as agents [1]. The case ServiceNow stresses is third-party agents. It says the passport gives a verifiable permissions record, so the platform does not have to rely on an agent's own claims about its permissions [6]. Jon Sigler, EVP and GM, AI Platform at ServiceNow, framed the product as closing a visibility gap: "As AI agents proliferate across the enterprise, visibility alone isn't enough. Enterprises need confidence that every AI agent is operating within its authorized scope," he said [11].
Inside its scope, ServiceNow says, an agent can operate freely [3]. In our view that is the limit to test. Drift responses fire when an agent leaves its scope. An agent manipulated into misusing authority its passport already grants does not leave it.
ServiceNow says the credentials rest on open specifications, meaning identity providers, governance products and enforcement points of any kind can both issue and verify them, for their own agents and for third-party agents alike [9]. The portion of the announcement published by Help Net Security does not name the specifications or give a ship date for the NVIDIA implementation.
What to watch
- Documentation naming the open specifications, and whether issuers and verifiers outside ServiceNow interoperate with them.
- A ship date and first customer deployments for the NVIDIA OpenShell and Sentry enforcement implementation.
- Independent testing of whether anything beyond the passport catches a prompt-injected agent acting inside its granted scope.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence25
- Adoption
- Insufficient
- Hype gap+40
- Incentives70
- Confidence60
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
ServiceNow announced AI Passport, an authority layer that gives enterprises runtime control over what AI agents and Model Context Protocol (MCP) servers are authorized to do at every point in a multi-agent workflow.
- [2]
AI Passport is a cryptographically signed credential that AI agents refer to at runtime, including in environments such as NVIDIA OpenShell, to verify their authorized scope.
- [3]
When an AI agent stays within its passport scope, it can operate freely.
- [4]
When an agent drifts, ServiceNow says graduated responses can be enforced: redirecting, restricting, or reconnecting the agent with proper context, including kill switch capabilities at machine speed if termination is needed.
- [5]
Each AI agent receives its own AI Passport, issued by the Super Intelligence Control Tower and grounded in the enterprise's live business context through the ServiceNow CMDB, GRC policies, and workflow approvals.
- [6]
For third-party AI agents especially, ServiceNow says AI Passport creates a verifiable permissions record rather than relying on the agent's own claims about its permissions.
- [7]
ServiceNow says the approach enables precision revocation and scoped delegation that narrows at each step for third-party agent governance, along with audit trails that include full compliance signatures.
- [8]
Credentials are verifiable at every step in multi-agent workflows.
- [9]
AI Passport is built on open specifications so that any governance product, identity provider, or enforcement point can issue and verify credentials for their own and third-party AI agents.
- [10]
ServiceNow announced NVIDIA as a design and launch partner, building a production-ready enforcement implementation using the NVIDIA Open Agent Safety Platform, which features NVIDIA OpenShell open-source secure runtime software and the NVIDIA Sentry security framework.
- [11]
"As AI agents proliferate across the enterprise, visibility alone isn't enough. Enterprises need confidence that every AI agent is operating within its authorized scope," said Jon Sigler, EVP and GM, AI Platform at ServiceNow.
- [12]
"As enterprises shift from experimenting with agents to running them in production, securing those agents demands enforcement at the infrastructure layer," said Itay Ozery, director of Product Marketing for Networking and Cybersecurity at NVIDIA.
- [13]
ServiceNow says that, unlike audit trails, which show what happened after the fact, AI Passport helps prevent AI agents from acting outside their scope by authorizing what they can do at runtime.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- helpnetsecurity.comServiceNow gives AI agents a passport, and limits on what they can do
1 article · October 15, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Entities
- ServiceNowFollow
- NvidiaFollow
- AI PassportFollow
- NVIDIA OpenShellFollow
- Jon SiglerFollow
- Itay OzeryFollow