Your organization is using Model Garden to maintain a collection of models in a single...

To centrally govern and restrict which AI models (from the Model Garden) can be used within an organization, Google Cloud provides a specific Organization Policy Constraint.9

According to Google Cloud Documentation (Vertex AI Organization Policy Constraints):

"The constraints/vertexai.allowedModels constraint allows you to define a list of allowed models that can be deployed or used within your organization. This includes Google first-party models, open-source models, and third-party models available in the Model Garden.10 By using this policy, you can prevent users from using unvetted or non-compliant models even if they have IAM permissions to use Vertex AI."

How it works:

You define an "Allowlist" of model IDs.

When a user attempts to deploy a model or call an API for a model not on the list, the request is blocked.

This is the most scalable and compliant way to manage AI governance.

Why other options are incorrect:

A is incorrect: Model Garden is a catalog; you cannot apply granular IAM permissions to individual entries within the public catalog itself in the same way you can restrict API usage via Org Policy.

B is incorrect: Auditing is reactive. The requirement is to ensure users can only access approved models (prevention).