ServiceNow AI Passport: Signed Permissions for AI Agents

AI agents are moving from demos into real business workflows. Companies now need a way to say exactly what each agent may do, and to enforce it while the agent runs. ServiceNow announced its answer on October 15: AI Passport, a cryptographically signed credential that every AI agent and Model Context Protocol (MCP) server carries. The credential is checked at each step of a workflow. Nvidia is the design and launch partner.

• Signed credential per agent: each agent and MCP server gets a passport stating its authorized scope, issued by ServiceNow's Super Intelligence Control Tower
• Built on open specifications: any governance product, identity provider or enforcement point can issue and verify passports
• Graduated enforcement: agents that drift can be redirected, restricted or reconnected, with a kill switch as the last resort
• Nvidia enforcement: a production implementation built on Nvidia's Open Agent Safety Platform, including OpenShell and the Sentry security framework

What Is an AI Agent Passport?

Think of a passport as an ID card and a visa in one. AI Passport tells every system an agent touches who the agent is and what it is allowed to do right now. The passport is grounded in data enterprises already maintain: the CMDB (the inventory of systems and assets), governance, risk and compliance policies, and existing workflow approvals.

Agents present the credential at runtime, including inside Nvidia OpenShell, to prove their scope before acting. ServiceNow also describes scoped delegation for third-party agents: permissions narrow at each step instead of passing along wholesale. It adds precision revocation and audit trails with compliance signatures. The company frames the passport as preventive, stopping out-of-scope actions before they happen instead of only recording them afterward.

How Does AI Passport Stop Agent Drift?

Enforcement is deliberately graduated, which keeps agents useful without being reckless:

• In scope: agents run freely, with no friction for approved work
• Drifting: the system can redirect the agent, restrict it, or reconnect it with the right context
• Out of bounds: a kill switch can stop the agent at machine speed

ServiceNow applies the same graduated idea to spending, with policies that move from alert to throttle to block.

"Enterprises need confidence that every AI agent is operating within its authorized scope," said Jon Sigler, ServiceNow's EVP and general manager of AI Platform.

Discovery, Shadow AI and Cost Controls

The announcement pairs the passport with tools to find agents in the first place. These include shadow AI detection, more than 400 out-of-the-box connectors for AI discovery, access mapping tied to real entitlements, and continuous observability of agent traces. ServiceNow also added AI cost management inside its IT asset management suite. Separately, it introduced an Operational Resilience offering for financial institutions. The company's 2026 Banking AI Maturity Index found that 100% of banks use AI, but only 24% have governance to match.

One customer shows the scale involved. LTM's CISO, Chandan Pani, described the control tower as "a single system of record" for the company's 1,500 AI agents, each of which has its own employee ID and a human manager.

Why Open Specifications Matter

Agent identity is quickly becoming a shared industry problem. Workday's Agent Passport and AWS's Strands Box sandbox approach it from different angles. ServiceNow's choice of open specifications means passports issued in one system can be checked in another. That is the kind of interoperability that lets zero-trust principles extend to software agents as well as people. Follow more agent-security developments in our AI security section.

Sources: Help Net Security — ServiceNow AI Passport — October 15, 2026; CIO Dive — ServiceNow boosts AI agent control — October 15, 2026; ServiceNow press release — October 15, 2026.