ServiceNow and Nvidia Launch AI Passport to Secure Enterprise AI Agents with Runtime Controls

ServiceNow and Nvidia Address Enterprise AI Agent Security Concerns

ServiceNow has partnered with Nvidia to launch AI Passport, a cryptographically signed credential system designed to govern AI agents across enterprise environments1

1

2

2

. The announcement addresses a critical challenge facing businesses: ensuring autonomous agents operate only within their authorized boundaries while scaling AI deployments from pilots to production. AI Passport gives every agent—whether from ServiceNow or third parties—a verifiable identity that defines permitted actions, accessible systems, and accountable owners3

3

. This matters because enterprises expanding agentic AI face new risks when they cannot identify which agents are running, what they're doing, or which systems they can access2

2

. The system represents a shift from passive monitoring to active enforcement, with Frank Dickson, principal analyst at Dickson Research, noting that "most agent governance today is a security camera" while "AI Passport is a bouncer at every door"2

2

.

Multi-Layer AI Governance Through Super Intelligence Control Tower

ServiceNow's renamed Super Intelligence Control Tower (SICT) serves as the policy hub for AI governance, working alongside Nvidia's Open Agent Safety Platform to enforce controls at multiple infrastructure layers1

1

4

4

. AI Passport draws on business context from ServiceNow's Configuration Management Database (CMDB), governance policies, and workflow approvals to establish enforceable limits on agent authority3

3

4

4

. When an agent stays within its passport scope, it operates freely, but deviations trigger defined actions including redirecting, restricting, or reconnecting the agent with proper context2

2

. In demonstrations, ServiceNow configured a customer service agent to process refunds below $25 automatically—an $18 refund proceeded, while a $40 request was rejected at runtime because it exceeded the threshold requiring human review3

3

4

4

. The approach supports graduated responses rather than binary on-off controls. Nenshad Bardoliwalla, Senior Vice President of AI Products at ServiceNow, explained: "Real governance is graduated. It's more like a dimmer switch. You monitor, you pause, you restrict, and eventually, if necessary, you deactivate"4

4

.

Runtime Enforcement and Hardware-Level Security Integration

The partnership integrates ServiceNow's enterprise policies with runtime enforcement through Nvidia's OpenShell secure runtime and BlueField data processing unit1

1

5

5

. This creates what ServiceNow calls "enforcement in silicon," designed to prevent overrides from compromised software by moving security beyond the software layer into runtime and hardware layers1

1

. When the Control Tower revokes an agent's passport, Nvidia BlueField enforces that revocation at machine speed before the agent can proceed1

1

. OpenShell can contain unauthorized tasks within a sandbox for later audit while allowing other approved agent actions to continue without restarting1

1

. Itay Ozery, Nvidia's director of product marketing for cybersecurity, emphasized the challenge as "making your enterprise policy meaningful at the point of execution"3

3

. The system uses open specifications, allowing other governance products, identity providers, and enforcement points to issue and verify credentials for their own and third-party AI agents2

2

4

4

.

Arc Agent and Enhanced Employee Automation Tools

ServiceNow and Nvidia also launched Arc, a personal long-running AI agent now generally available on Mac with Windows support coming1

1

5

5

. Built for Nvidia's RTX Spark devices, Arc handles everyday work tasks relying on productivity tools like email and Microsoft Teams, running locally on desktops or within Nvidia OpenShell1

1

. The agent builds memory and skills over time, handing complex tasks to frontier models only when necessary while keeping data as local as possible1

1

. ServiceNow's third-generation EmployeeWorks now enables employees to build Productivity Assistants and automate routine tasks without coding3

3

. ToolSmith, a new agentic developer environment, reduces tool building "from days to minutes" by letting developers create automations through natural language descriptions2

2

3

3

. Trained on over 75,000 APIs and optimized for reliability, ToolSmith can plan integrations, retrieve logs, suggest fixes, and support testing2

2

3

3

.

Expanded Observability and Cost Governance Capabilities

The Super Intelligence Control Tower now includes shadow AI detection to locate unapproved agents on networks, more than 400 connectors for AI discovery across platforms and vendors, access mapping, and continuous enterprise AI observability2

2

3

3

. New cost governance features track spending by user, department, and runtime while connecting consumption, cost, quality, and realized user value to help teams optimize AI use2

2

3

3

. AI Cost Management consolidates spending on frontier models, AI embedded in software services, hyperscalers, and data center GPUs3

3

. Bardoliwalla positioned ServiceNow as uniquely capable of bringing discovery, observability, governance, compliance, security, and cost measurement together holistically, stating: "We are the only product that brings all five of those pieces together holistically, connecting the dots in the same platform"4

4

. AI Passport, shadow AI detection, and AI Cost Management are expected to be available by year end2

2

. The announcement comes as consumer trust in AI agents varies significantly—while 56% will let agents search and compare products, only 35% will allow access to saved payment credentials, according to PYMNTS Intelligence research5

5

.