ServiceNow launched AI Passport on October 15 — a cryptographically signed runtime credential that tells every system in a multi-agent workflow exactly what an AI agent is authorised to do, and stops it when it drifts. NVIDIA is the launch partner, building enforcement into its Open Agent Safety Platform using hardware-backed security that can quarantine a rogue agent in milliseconds. The product arrives at a moment when enterprises are deploying AI agents faster than they can govern them: ServiceNow's own research found that all banks surveyed use AI, but just 24% have testing, auditing, and risk processes in place.
Why this matters now
Most enterprise AI governance today works like a security camera. It records what an agent did so someone can review the footage later. That approach was tolerable when AI agents were experimental tools with limited access. It falls apart when agents are making purchasing decisions, processing customer refunds, or accessing production databases autonomously.
The shift to agentic AI — where software agents plan, decide, and execute multi-step workflows without waiting for human approval — has created an identity crisis. Who is this agent? What is it allowed to touch? Who authorised it? And critically: can you stop it before it does something it shouldn't?
This is the problem ServiceNow is targeting, and the timing is not accidental. The Five Eyes intelligence alliance published guidance earlier this year explicitly recommending that every AI agent be treated as a distinct cryptographic identity. Australia's own APRA wrote to the entire financial sector warning that governance has not kept pace with AI adoption. And the IETF is drafting formal standards for AI agent identity credentials through its WIMSE working group. AI Passport sits at the intersection of all three trends.
What AI Passport actually does
Each AI agent — whether built on ServiceNow or from a third party — receives its own cryptographically signed credential issued by ServiceNow's Super Intelligence Control Tower. The credential is grounded in the enterprise's live business context: its Configuration Management Database (CMDB), governance and compliance policies, and workflow approvals.
The critical difference from traditional access control is that the credential is checked at runtime, at every step of a multi-agent workflow. When an agent operates within its passport scope, it moves freely. When it drifts beyond its authorised boundaries, graduated responses kick in — redirecting the agent, restricting its access, or triggering a kill switch at machine speed.
"Real governance is graduated," said Nenshad Bardoliwalla, SVP of AI Product Management at ServiceNow. "It's more like a dimmer switch. You monitor, you pause, you restrict, and eventually, if necessary, you deactivate."
For third-party agents — the ones your business didn't build but relies on — this is particularly significant. Instead of trusting the agent's own claims about its permissions, AI Passport creates a verifiable, externally issued permissions record. The credential uses open specifications, meaning other governance products, identity providers, and enforcement systems can issue and verify passports too.
A bouncer at every door, not a security camera
The strongest framing came from Frank Dickson, principal analyst at Dickson Research. "Most agent governance today is a security camera. It records what happened so someone can review the tape later," he told CIO. "AI Passport is a bouncer at every door."
Dickson's point cuts to the core of why this matters. Audit trails are necessary but insufficient. When Microsoft's Delegate agents corrupted documents earlier this year, the audit trail showed exactly what went wrong — after the damage was done. ServiceNow's "Shift Zero" approach aims to prevent out-of-scope actions before they happen.
But Dickson also raised the sharpest criticism: "A passport is only as trustworthy as the office that issues it. Here, the issuing office is ServiceNow's CMDB, its GRC policies, and its workflow approvals. If that data is stale, the passport is stale." He also noted that enforcement is still being built with NVIDIA, and an open specification only matters when others adopt it.
NVIDIA's contribution is the hardware-backed enforcement layer. Its Open Agent Safety Platform, launched in September, includes OpenShell — an open-source secure runtime — and Sentry, an out-of-band watchdog running on BlueField-4 DPUs that monitors agent behaviour independently and can quarantine agents in milliseconds. "By implementing AI Passport with the NVIDIA Open Agent Safety Platform, ServiceNow customers can continuously verify and enforce what every agent is authorized to do — at machine speed and at runtime," said Itay Ozery, NVIDIA's Director of Product Marketing for Networking and Cybersecurity.
The governance gap — and why Australia should pay attention
The 24% figure from ServiceNow's banking study is stark, but it maps onto a global pattern. Governance consistently lags adoption across industries, and Australia is no exception.
The Australian Signals Directorate's cyber.gov.au guidance on agentic AI adoption now explicitly recommends that organisations "construct each agent as a distinct principal, a cryptographically anchored identity with its own unique keys or certificates." It calls for mutual TLS on all inter-agent API calls, a trusted registry binding identities to authorised roles, and role-based identity management limiting agents to the minimum scope required. This is essentially the same architecture AI Passport implements.
APRA has been even more direct, warning that few regulated entities have operationalised AI governance in practice and flagging gaps in post-deployment monitoring, model behaviour tracking, and change management. In August, APRA and ASIC jointly warned that frontier AI awareness must turn to action, with board-level escalation and resilience controls expected before a crisis, not during one.
For Australian businesses — especially those in financial services, insurance, or superannuation — the regulatory trajectory is clear. Agent governance is shifting from recommended to expected. Products like AI Passport represent what compliance will look like in practice: cryptographic identity, runtime verification, scoped permissions, and auditable enforcement.
What this means for your business
You don't need to be a ServiceNow customer to take the lesson here. The principles behind AI Passport apply to any business deploying AI agents:
Treat every agent as an identity. Not a tool, not an integration — an identity with its own credentials, permissions, and audit trail. The days of giving an AI agent a shared API key and hoping for the best are ending.
Scope permissions narrowly and enforce them at runtime. Static access lists aren't enough when agents can reason their way into unexpected actions. Your agents need permissions that are checked continuously, not just at login.
Own the governance layer before someone else does. ServiceNow CEO Bill McDermott's ambition is explicit: "We are the AI agent of the agents. We manage everyone else's agents." Whether that pitch wins or not, the underlying point is sound — whoever controls agent governance controls the enterprise. If you're using agents from multiple vendors, you need a single governance view.
Budget for governance, not just deployment. As Dickson noted, "An agent should carry its permissions, its manager, and its budget wherever it goes. But until the brakes ship, that budget is a decision a person still has to make."
What to watch
AI Passport, along with ServiceNow's shadow AI detection and AI Cost Management, is expected in Q4 2026. NVIDIA's enforcement layer is being built in parallel. The real test will be whether the open specifications attract adoption beyond ServiceNow's own ecosystem — identity providers like Okta and Google Cloud are building competing approaches to agent identity, and the IETF's WIMSE working group is drafting formal standards that could unify or fragment the market.
The broader trend is unmistakable. AI agents are becoming a permanent part of how businesses operate, and the infrastructure for governing them is being built right now. The companies that treat agent governance as a day-one requirement — not a future compliance checkbox — will be the ones that scale AI safely. The ones that don't will learn the lesson the hard way, one rogue agent at a time.
At Heygentic, we've been tracking this shift from Cloudflare and Stripe building economic infrastructure for agent transactions to IBM's multi-agent control plane. ServiceNow's AI Passport is the clearest signal yet that agent identity isn't a niche security concern — it's the foundation of the entire agentic enterprise.
Sources
- ServiceNow advances its operational harness for enterprise AI with AI Passport — ServiceNow Newsroom
- ServiceNow makes it easier to build AI agents, gives them verifiable identities — CIO (Lynn Greiner)
- ServiceNow announces AI Passport — bringing authorisation, business context and a kill switch together — Diginomica (Derek du Preez)
- ServiceNow Gives AI Agents Signed Runtime Credentials — AI.info
- NVIDIA Launches Open Agent Safety Platform to Secure Agents From Testing to Deployment — NVIDIA Newsroom
- 2026 Enterprise AI Maturity Index for Banking — ServiceNow / ThoughtLab
- Careful adoption of agentic AI services — Australian Signals Directorate (cyber.gov.au)
- APRA Letter to Industry on Artificial Intelligence — APRA
- Your company's AI could delete everything in 9 seconds — Fortune
