AI governance

AI governance extends Unity Catalog to the AI your organization uses. Unity AI Gateway is the control plane for AI agents: it routes every model and MCP request, enforces rate limits and cost controls, applies service policies, and records usage and cost for observability — across any model provider and any coding agent. Unity Catalog governs the assets behind it with the same access control, lineage, and audit model that protects your data.

Azure Databricks governs AI across three dimensions:

• Access to AI assets: Unity Catalog manages models, MCP servers, and functions as securable objects, governed with standard Unity Catalog privileges and ABAC grant policies.
• Traffic and cost: Unity AI Gateway — the agent control plane — routes every model and MCP request, enforces rate limits and cost controls, and records usage, for any provider.
• Request and response content: Service policies allow, deny, or require approval for each interaction, based on its content and on who is calling.

If you are new to AI governance on Azure Databricks, see Get started with AI governance for an end-to-end setup path.

AI asset governance with Unity Catalog

Unity Catalog manages AI assets as securable objects. You can grant and revoke access to the following AI assets using standard Unity Catalog privileges:

• Models: Registered ML models in Unity Catalog, including hosted foundation models, which are Databricks-hosted foundation models available through Foundation Model APIs. See Manage model lifecycle.
• MCP servers: Registered as Unity Catalog securables (MCP Services), with tool filtering and service policies. See Connect agents to third-party tools with MCP Services.
• Agents: Registered as Unity Catalog securables and governed alongside your tables, models, and functions.
• Connections: Unity Catalog HTTP connections used to access external APIs and MCP servers. See HTTP connections.
• Functions: Unity Catalog functions used as agent tools or for data transformations. See Create AI agent tools using Unity Catalog functions.

Traffic and cost governance with Unity AI Gateway

Important

This feature is in Beta. Account admins can manage access to this feature from the account console Previews page. See Manage Azure Databricks previews.

Unity AI Gateway is the agent control plane: it routes traffic to the model services and MCP services your organization uses, enforces rate limits and cost controls, and records usage from a central location:

• Model services: Manage access to both Azure Databricks-hosted and external-provider models (such as OpenAI, Anthropic, and Google), enforce rate limits, track usage and cost across providers, and set up traffic splitting and fallbacks.
• MCP services: Manage access to managed, external, and custom MCP services alongside your model services, with usage tracking and rate limits.

Model services are Unity Catalog securables, so they're accessible from any workspace in the account, unless catalog-workspace bindings restrict them. To track usage and analyze cost across providers, see Model usage for Unity AI Gateway services and Monitor Unity AI Gateway cost.

Request and response content governance with service policies

Important

This feature is in Beta. Account admins can manage access to this feature from the account console Previews page. See Manage Azure Databricks previews.

Unity Catalog grants determine whether a principal can call an AI service. Service policies govern how that interaction proceeds, based on the content of the request and response and on who is making the call. This matters most when agents act on behalf of users and reach external systems.

A service policy is a type of attribute-based access control (ABAC) policy scoped to AI services. You can allow, deny, or require human approval for an interaction based on its content — for example, to block personally identifiable information (PII) or deny an out-of-policy tool call.

See Service policies for AI securables.

In this section

The following topics help you get started with AI governance on Azure Databricks.

Topic	Description
Get started with AI governance	An end-to-end path for administrators to govern access to AI assets, traffic and cost, and the content of requests and responses, with guidance for choosing the right approach for each requirement.
Service policies for AI securables	How service policies govern the content of requests and responses to AI services, using built-in guardrails and custom policies.
Create and attach a service policy	How to write a service policy function and attach it to an MCP Service or Model Service.
Unity AI Gateway	How Unity AI Gateway governs model services and MCP services, including permissions, rate limits, usage, and cost.