ServiceNow Goes All In on Governing Every AI Agent

In October 2026, ServiceNow went all in on AI agent governance. Across five announcements in a single month, it extended its reach from its own agents to everyone's, betting that the control plane over the agents matters more than the agents themselves.

The headline is AI Passport: a cryptographically signed, scoped credential for every agent and MCP server, checked at each step of a workflow, for ServiceNow and third-party agents alike. Around it sit shadow AI discovery, kill switches, AI cost management, and an 11-member Control Tower Alliance that includes Anthropic, AWS, Cisco, NVIDIA, Okta, and OpenAI.

The differentiator is business context. An identity provider knows who an agent is. ServiceNow claims to know what the agent should be doing right now, scoping its authority from the CMDB, GRC policies, and workflow approvals. The weakness is enforcement: beyond its own platform, ServiceNow decides and partners such as Okta and NVIDIA block. And most of it is not generally available yet. AI Passport is expected in Q4 2026, and most Alliance integrations are still commitments rather than shipping code.

The direction is right, and the architecture is credible. CISOs should treat ServiceNow as a strong candidate for the AI governance system of record, especially where it already runs the CMDB and GRC, while keeping enforcement anchored in the identity and security tools they already trust.

The brief covers the five announcements, how ServiceNow stacks up against Microsoft, Okta, Salesforce, CrowdStrike, and Palo Alto Networks, the questions to put to ServiceNow, and seven recommendations for CISOs. The short version: act on the principles now and the products later. Inventory every agent, give each one an identity and a named human owner, and test your kill switch end to end.