AI policy

AI policies define who can view, edit, or restrict access to AI assets in Atlan. They let admins control changes to AI models and applications, ensuring that only authorized users can create, modify, or manage AI assets. They also govern access to AI governance properties, ensuring compliance and proper management of AI resources.

Use AI policies to implement fine-grained governance over AI assets, ensuring consistency and compliance across your AI ecosystem. AI policies govern access to:

• AI models: creating, updating, and managing AI models
• AI applications: creating, updating, and managing AI applications
• Tags and terms: adding, removing, and managing AI asset associations
• Governance properties: certification status and ownership of AI assets
• API operations: programmatic access to AI assets

AI policies don't govern data access (the actual querying and retrieval of data). They also don't govern Connection Admin privileges, which grant automatic access to all assets within a connection. Additionally, they don't govern bootstrap permissions, which are the role-based baseline permissions that apply outside of personas.

Access

AI policies aren't enabled by default. To use them, you must be an admin user with permissions to define personas and governance settings.

You can access this in Governance → Access Control → Personas, then open a persona (or create a new one), go to the Policies tab, and click New Policy → AI Policy. Once saved, the policy applies to all users and groups linked to that persona.

Name

Specifies a unique name to identify the AI policy in Atlan. This name must briefly describe the purpose or scope of the policy.

Example:

test

Select AI assets

Choose the AI assets on which this policy is applied. You can select specific AI models, AI applications, or apply the policy to all AI assets.

Example:

All AI models and applications

Configure permissions

Define the level of control that personas have over AI assets. Permissions can be combined to give broad or fine-grained access.

Edit

Controls the ability to modify AI asset properties and configurations.

Assets

Controls access to the core properties of AI assets.

• Create: create new AI models or applications.
• Read: view AI asset metadata such as names, descriptions, and configurations.
• Update: edit AI asset properties like descriptions, owners, and classifications.
• Delete: remove AI models or applications.

Governance

Manages governance properties and associations for AI assets.

• Add tags: add Atlan tags to AI assets for classification and organization.
• Add terms: associate glossary terms with AI assets to link business definitions.
• Update custom metadata values: edit custom metadata structures applied to AI assets.
• Remove tags: remove tags from AI assets.
• Remove terms: disassociate glossary terms from AI assets.

Example:

• Add the tag Production to an AI model.
• Add the glossary term Customer Analytics to an AI application.
• Update the "Model Version" field in a custom metadata template.