| Deployment | Deployed inside your own cloud (AWS, Azure, GCP or on-premises). Model, tool and data flows are set by your environment and the providers you approve. | A ServiceNow-hosted platform. It governs AI deployed across other systems, so the platform boundary is not the same as the governance boundary. |
|---|
| Agent and AI coverage | Supports agent deployment with LangGraph. | Discovers AI assets across the enterprise — including systems beyond ServiceNow — through 30 integrations spanning AWS, Google Cloud, Azure, SAP, Oracle and Workday, extending to non-human identities and connected OT/IoT devices. |
|---|
| Evidence | Release evidence linking the evaluated build to its production approval, including the commit and image digest. | Continuous monitoring with live metrics and alerts, plus runtime observability of agent behaviour through the Traceloop acquisition: how agents reason, where they decide, and when to course-correct. |
|---|
| Identity and enforcement | Provisions agent identities through Microsoft Entra. Scoped credentials and a policy gateway check each tool call before execution. | Identity access governance extended to hyperscaler AI environments and connected devices through Veza's access graph, with scoped permissions and least-privilege enforcement; agents that exceed their permissions can be detected and shut down in real time. |
|---|
| MCP control | The policy gateway sits in front of MCP tool calls and can block or pause them before execution. | A new control for customer Model Context Protocol transactions provides real-time controls for agentic workloads, with governance, observability and security across third-party AI systems. |
|---|
| Risk frameworks | Control mappings to SR 26-2, FFIEC, NAIC, HIPAA and the EU AI Act. SR 26-2 excludes generative and agentic AI, so the SR 26-2 entries are Moring's suggested practices for agent workflows rather than requirements under the guidance. A starting crosswalk, not a certification. | AI-driven risk assessment across agents, models, data sets, prompts and classic machine learning, with five risk frameworks aligned to NIST and EU AI Act standards. |
|---|
| Cost and ROI | Agent-specific model-gateway keys support budget controls. The Stop control blocks gateway access and new plane actions, without cancelling work already in flight. | Cost tracking and ROI dashboards for AI spend, aimed at runaway model costs and at showing where AI returns value. |
|---|
| Release status | Production capability, deployed per customer. | The May 2026 capabilities were announced with general availability expected in August 2026. Confirm the current status of each capability you rely on during evaluation. |
|---|