Expected: ServiceNow Plans Q4 AI Passport for Runtime Agent Permissions
The planned signed credential would let enterprises verify agent and MCP-server authority during workflows, with availability still expected later in 2026.
Edited by Tyronne Panaino
ServiceNow announced AI Passport on October 15 as a planned authority layer for AI agents and Model Context Protocol servers. Its official announcement says the product is expected in the fourth quarter of 2026, so this article treats the release and its promised capabilities as an expectation rather than a shipped fact.
If delivered as described, the signed credential would give an enterprise a runtime check on what an agent is allowed to do during a multi-agent workflow. That matters to teams moving from agent inventories and after-the-fact audit logs toward controls that can restrict an action while it is happening. ServiceNow has described the design, but no independently testable release, final launch date or cross-platform interoperability result was fetched.
What is known
ServiceNow describes AI Passport as a cryptographically signed credential that an agent refers to at runtime to verify its authorized scope. The credential is meant to be grounded in enterprise context from the company's configuration, governance and workflow-approval systems. The stated aim is to give first- and third-party agents a verifiable permissions record instead of accepting an agent's own representation of its authority.
The company says the design uses open specifications so governance products, identity providers and enforcement points can issue and verify credentials. It also describes scoped delegation and precise revocation, with authority narrowing as a workflow moves through different agents. Those are product-design claims from ServiceNow, not independently demonstrated interoperability guarantees.
The planned response model is graduated. ServiceNow says an agent that moves outside its scope could be redirected, restricted or reconnected with appropriate context, with termination available when needed. Its Super Intelligence Control Tower is intended to retain the audit record behind those actions. The announcement establishes the intended control flow; it does not show how reliably the system identifies drift or how quickly every enforcement point responds.
NVIDIA is named as a design and launch partner. ServiceNow says the companies are building an enforcement implementation using the NVIDIA Open Agent Safety Platform, including its OpenShell runtime and Sentry framework. The announcement does not provide an independent conformance test, supported-version matrix or evidence that the partner implementation is generally available today.
How the credential model changes the control point
The key proposed shift is from recording an agent's actions after execution to checking delegated authority during execution. In practice, that could let an organization separate three questions: which identity requested an action, what scope was delegated and which enforcement point approved or blocked it. This is an inference from the announced credential and audit design, not a measured customer outcome.
Open specifications could also matter for mixed environments where agents come from more than one vendor. A common credential format would be useful only if independent issuers and enforcement systems implement the same semantics for scope, delegation and revocation. ServiceNow says the design is open, but the fetched evidence does not identify a public specification, governance body, compatibility suite or adoption list.
What remains prospective
ServiceNow says AI Passport, shadow-AI detection and AI Cost Management are expected to become available in Q4 2026. By contrast, the company's Operational Resilience solution for financial services is described as available now. Keeping those timelines separate is important: the announcement is not evidence that AI Passport can already be purchased, deployed or tested by every customer.
The press release also contains a forward-looking-statement warning. Availability, capabilities and expected benefits may change, and the company identifies execution delays, regulatory changes and uncertain demand as possible factors. That caution reinforces the expectation label used here.
What would confirm it
A generally available product page with a dated release, public technical specification and supported integration list would confirm that the planned credential has moved beyond announcement. Independent tests should then check whether third-party agents can present and narrow scopes consistently, whether revocation reaches every enforcement point and whether audit records match the actions allowed or blocked.
A second checkpoint would be evidence from customers using the system under realistic multi-agent workflows. Useful proof would separate inventory visibility from actual runtime enforcement and show how the control behaves when credentials are stale, scopes conflict or a dependent service is unavailable. None of that evidence was fetched for this report.
Status
Expectation. ServiceNow has announced AI Passport and a Q4 2026 availability target; internal confidence is medium because the schedule and capabilities are first-party and remain forward-looking.
Sources
Update note: Last reviewed 2026-10-15. We will revise this post when ServiceNow publishes a final release date, technical specification, supported integrations or independently reviewable deployment evidence.
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.