ServiceNow AI Passport: Your Agents Need Cryptographic Credentials Now

ServiceNow announced AI Passport on October 15, an enterprise system that issues AI agents cryptographically signed credentials encoding identity, permitted actions, accessible systems and accountable owner, verified at runtime by the AI Control Tower before every action. ServiceNow SVP Nenshad Bardoliwalla said "Without that baseline, every guardrail is just advisory," and the company cited figures showing 100% of banks run AI agents while only 24% have governance frameworks and 45.1% of organizations have basic audit logging for agent actions. Built on open specifications with NVIDIA as design and launch partner contributing its Open Agent Safety Platform (OpenShell runtime and Sentry security framework), AI Passport reaches general availability in Q4 2026.

100% of banks now run AI agents. 24% have governance frameworks to match. The other 76% are deploying autonomous systems that can approve transactions, trigger workflows, and touch production data — with no enforceable limit on what they actually do. ServiceNow’s AI Passport https://newsroom.servicenow.com/press-releases/details/2026/ServiceNow-advances-its-operational-harness-for-enterprise-AI-with-AI-Passport-and-expanded-governance-capabilities/default.aspx , announced October 15, is the first enterprise system to replace advisory policy with cryptographically verified credentials that agents must present before every action. The Problem With Policy-Based Control Most teams govern AI agents the same way they govern employees: write a policy, assume it gets followed, audit violations after the fact. That works when the actor can read the rulebook and reason about intent. It does not work when the actor is an autonomous agent running 10,000 requests per minute. ServiceNow SVP Nenshad Bardoliwalla put it bluntly: “Without that baseline, every guardrail is just advisory.” An agent authorized to approve refunds up to $25 has no mechanical barrier from approving $4,000 if its instructions are ambiguous, its context is manipulated, or it simply drifts. Policy says stop. Nothing stops it. This is the gap the market finally has a number for: 45.1% of organizations have basic audit logging for agent actions https://futurumgroup.com/insights/servicenows-ai-passport-closes-the-agentic-governance-gap/ . Fewer than half. The rest find out what their agents did by checking damage reports. How AI Passport Works AI Passport issues each agent a cryptographically signed credential at provisioning time. The credential encodes the agent’s identity, permitted actions, accessible systems, and accountable owner — grounded in the organization’s Configuration Management Database, GRC policies, and workflow approvals. It is not a permission slip the agent carries in memory. It is a verifiable proof that any enforcement point can check. At runtime, before every action, the agent presents this credential. The AI Control Tower verifies it cryptographically. If the action falls within scope, it proceeds. If it exceeds scope, enforcement kicks in on a graduated scale — not a binary pass/fail. ServiceNow describes this as a dimmer switch: - Monitoring — action logged, no interruption - Pausing — agent waits for human confirmation - Restricting — access narrowed, execution limited - Deactivating — kill switch at machine speed ServiceNow demonstrated this with a live scenario: a third-party agent configured for sub-$25 refund approvals processed an $18 request automatically. A $40 request was rejected at runtime. Not flagged for review — rejected. That is the difference between policy and enforcement. NVIDIA Closes the Gap Between Spec and Production A governance system that runs entirely inside one platform has a ceiling. ServiceNow built AI Passport on open specifications — any identity provider, governance product, or enforcement point can issue and verify credentials for first- and third-party agents. NVIDIA is the design and launch partner, contributing its Open Agent Safety Platform: OpenShell secure runtime and Sentry security framework . NVIDIA’s Itay Ozery described the core challenge: “making your enterprise policy meaningful at the point of execution.” For teams running on NVIDIA infrastructure, enforcement is native from day one. For everyone else, the open spec approach https://forkast.news/servicenow-bets-the-credential-is-the-governance-layer-not-the-policy/ means you are not waiting for ServiceNow to build every integration — any identity provider can participate. What Developers Need to Do Before Q4 2026 AI Passport reaches general availability in Q4 2026. Teams that wait for GA to start thinking about agent identity will be retrofitting governance into systems that were not designed for it. Four things to get right now: 1. Audit your current agent permissions. What can each agent in your system actually do? Not what the policy says — what would nothing technically stop it from doing? That gap is your risk surface. 2. Map actions to business context. Which agents touch which systems? What data scope do they need? AI Passport credentials are grounded in CMDB and GRC data. You need that mapping before you can generate meaningful credentials. 3. Design for graduated enforcement. Agents should degrade gracefully when restricted, not crash. If your agent has no handling for “action paused pending approval,” you have a fragile pipeline. 4. Watch the NVIDIA Open Agent Safety Platform spec. OpenShell and Sentry https://siliconangle.com/2026/10/15/servicenow-adds-ai-agent-controls-and-employee-automation-tools/ are the enforcement runtime for NVIDIA infrastructure. Understanding the spec now means your agents are ready when enforcement goes live. Agent Identity Is Becoming Infrastructure Cross-agent governance emerged as a top enterprise concern for the first time in 2026 surveys — 7.1% of enterprises naming it as their biggest agentic AI challenge, up from near zero. That number will grow as multi-agent workflows move from experiments to production systems. The protocol stack is settling: OAuth 2.1 for browser-based agents, JWT bearer assertions for service-to-service flows, and now cryptographic credentials for runtime scope enforcement. The teams that figure out agent identity before it becomes mandatory will have a significant advantage when regulators and auditors ask: “What did your agents do, and what were they authorized to do?” Right now most of us cannot answer that question. AI Passport is one serious attempt to make the answer possible.