Policies & Policy Editor

Policies are the core building block of Metatate. A policy defines governance rules for a set of tables and columns -- what data is sensitive, how it should be masked, who can access it, and how AI tools should handle it.

Policy List

The Policies page shows all policies in your application, each displaying its name, scope, and current status.

Screenshot needed

the policy list page

Status	Meaning
Draft	Created or edited but not yet deployed. Changes are local only.
Deployed	Applied to governance tables and visible to MCP tools.

You can filter, search, and sort the list to find policies quickly.

Creating a Policy

There are two ways to create a new policy:

AI-Assisted (Natural Language)

Type a plain-English description of what you want to govern, and Snowflake Cortex generates the policy YAML for you. For example:

"Mask all SSN and email columns in the customers table. Only the compliance team should see unmasked values."

Cortex runs entirely inside Snowflake -- zero data egress. The generated YAML appears in the editor for you to review and refine before saving.

Screenshot needed

the natural language input

Manual YAML

Open the policy editor and write YAML directly. This is useful for precise control or when duplicating an existing policy with modifications.

Policy Editor

The editor has two views:

• YAML View -- Edit the raw policy definition with syntax highlighting.
• Visual Review -- A read-friendly summary of what the policy does, organized by instruction type.

Screenshot needed

the policy editor with both views

Policy YAML Structure

Every policy follows this structure:

metadata:
  name: customer-pii-protection
  description: Protect PII in the customers table
  version: 1

scope:
  database: ANALYTICS
  schema: PUBLIC
  tables:
    - CUSTOMERS

instructions:
  - type: classification
    columns: [SSN, EMAIL]
    sensitivity: high
    pii: true

  - type: masking
    columns: [SSN]
    strategy: full
    except_roles: [COMPLIANCE_ADMIN]

Metadata

• name -- Unique identifier for the policy.
• description -- Human-readable summary.
• version -- Incremented on each edit.

Scope

Defines which tables (and optionally columns) the policy targets. References tables from your DATABASE_CATALOG.

Instruction Types

Type	Purpose
classification	Label columns with sensitivity level and PII status
masking	Define masking strategies and role-based exceptions
usage_guidance	Provide human- and AI-readable guidance on how data should be used
ai_governance	Rules specifically for AI/LLM access to data
retention	Data retention periods and archival rules
access_control	Role-based access rules beyond masking

A single policy can contain multiple instructions of different types.

Managing Policies

• Save -- Persists the policy as a draft. Does not affect deployed state.
• Edit -- Reopen any policy in the editor. Editing a deployed policy creates a new draft version.
• Delete -- Removes the policy. If it was deployed, you must redeploy to remove its governance rules.

Next Steps

Once your policies are ready, head to Deployments to plan and apply them.