ServiceNow launches AI Passport to control AI agent permissions and expands governance platform

On Thursday, 15th, ServiceNow announced a new digital credential called AI Passport, which acts as a passport for artificial intelligence agents. The tool allows companies to verify, during task execution, which tools, data, and actions each agent is authorized to access. Nvidia is the first partner in this initiative, which is expected to launch in the fourth quarter of 2026.

Why the passport matters

This initiative addresses a growing challenge: as more AI agents gain autonomy to choose tools and execute action sequences, companies need to know exactly who is operating, who authorized them, and how far they can go. Without this visibility, scale becomes a risk. “You can’t govern what you can’t identify. Companies need to have this governance,” says Nenshad Bardoliwalla, Vice President of Product Management at ServiceNow.

How AI Passport works

AI Passport is a cryptographically signed credential that ensures its authenticity. It records the authorized scope of each agent based on the organization’s policies, systems, and approvals. During execution, the system monitors whether the agent stays within the established limits. If the agent exceeds these limits, the platform can apply gradual measures: redirect the activity, restrict access, or halt execution completely.

Additionally, the mechanism allows for revoking specific permissions and limiting what an agent can delegate to another, which is crucial in environments with multiple interconnected systems. ServiceNow’s AI Control Tower, a corporate oversight platform, collects all action history and who authorized them, creating an auditable record.

Nvidia as a strategic partner

Nvidia was chosen as the first launch partner for AI Passport. The solution was built on open specifications, allowing other identity, governance, and security providers to issue or verify credentials. “The challenge with agents is that they can choose their own tools, and we have this gap that we’re working to close. We need both companies to help our customers use agents with confidence,” says Itay Ozery, Director of Product Marketing for Networking at Nvidia.

AI Passport also integrates with Model Context Protocol (MCP) servers, which connect agents to tools and information sources, ensuring that permissions are verified at every stage of the work, even when third-party components are involved.

Detecting unauthorized AI (shadow AI)

Another new feature is the ability to identify the use of AI without approval or knowledge of the responsible teams, known as shadow AI. ServiceNow expects to identify unauthorized agents operating on the network and expand the inventory of AI systems.

The expansion of AI Control Tower includes over 400 ready-to-use connectors to discover applications across different platforms and vendors, as well as map access and continuously monitor agent execution logs. This information helps define operational boundaries, linking each agent to the resources it can access and the rules it must follow.

AI cost management

The launch also includes AI Cost Management, a tool to monitor expenses related to AI models, resources embedded in enterprise software, cloud services, and graphics processors in data centers. Integration with AI Control Tower enables alerts about potential budget overruns and helps teams assess whether technology use delivers expected results.

Availability

All three new features—AI Passport, shadow AI detection, and AI Cost Management—are expected to launch in the fourth quarter of 2026.

The initiative reflects the growing need for governance in a scenario where AI agents operate increasingly autonomously. By combining permission verification, unauthorized use detection, and cost control, ServiceNow aims to provide companies with a solid foundation for adopting artificial intelligence at scale.