# Define access to models on Vertex AI

Facts read 10 Nov 2026. Page reviewed 16 Nov 2026.

Define access to models on Vertex AI is a Google Cloud guardrail written in organization Policy Service. Defines the set of models and features, such as predict, deploy and tune actions, that can be used in Vertex AI APIs. No documented way to test it before enforcement was found on 10 Nov 2026.

```yaml
name: organizations/ORGANIZATION_ID/policies/vertexai.allowedModels
spec:
 rules:
 values:
  deniedValues:
  - publishers/meta/models/llama3:deploy
  - publishers/google/models/gemini-3.5-flash:tune
  - publishers/hf-google/models/gemma-2b:deploy
```

- **Blocks:** Defines the set of models and features, such as predict, deploy and tune actions, that can be used in Vertex AI APIs. (legacy managed constraint, list.)
- **Breaks:** Google documents that the Model Garden policy only applies to models in Model Garden, not models registered in the Model Registry, and that a policy cannot exceed 500 allowed and denied values. ([source](https://cloud.google.com/vertex-ai/generative-ai/docs/control-model-access))
- **Test before applying:** As for every Organization Policy Service guardrail: [how to test before enforcing](/mechanisms/org-policy/)
- **Mechanism:** [Organization Policy Service (managed and legacy constraints)](/mechanisms/org-policy/): Resources in the organization, folder or project where the policy is set and its descendants; Google says application of new policies is usually not retroactive.
- **Taken from:** [provider documentation](https://cloud.google.com/vertex-ai/generative-ai/docs/control-model-access), provider documentation

## What the source says

"Model Garden's organization policy lets you centrally control the models your users can access and the actions they can take." (read 10 Nov 2026).

On what it breaks: "The Model Garden organization policy only applies to models in Model Garden."

## Related guardrails

- [Region deny that allows Bedrock cross-Region inference](/guardrails/aws-scp-region-deny-with-bedrock-cris/): AWS, Service control policies (SCPs), ai services
- [Allow only approved Amazon Bedrock models](/guardrails/aws-scp-bedrock-approved-models-only/): AWS, Service control policies (SCPs), ai services
- [Deny Amazon Bedrock API keys](/guardrails/aws-scp-deny-bedrock-api-keys/): AWS, Service control policies (SCPs), ai services
- [Restrict public IP access on new Vertex AI Workbench instances](/guardrails/gcp-orgpolicy-ainotebooks-restrict-public-ip/): Google Cloud, Organization Policy Service (managed and legacy constraints), ai services
- [Disable root access on new Vertex AI Workbench notebooks and instances](/guardrails/gcp-orgpolicy-ainotebooks-disable-root-access/): Google Cloud, Organization Policy Service (managed and legacy constraints), ai services
- [Disable Grounding with Google Search in generative AI APIs](/guardrails/gcp-orgpolicy-vertexai-disable-google-search-grounding/): Google Cloud, Organization Policy Service (managed and legacy constraints), ai services

## Related pages

- [Disable Grounding with Google Search in generative AI APIs (Google Cloud Organization Policy Service)](/guardrails/gcp-orgpolicy-vertexai-disable-google-search-grounding/)
- [Restrict egress domains for data connectors (Google Cloud Organization Policy Service)](/guardrails/gcp-orgpolicy-discoveryengine-allowed-egress-fqdns/)
- [Shared reservations owner projects (Google Cloud Organization Policy Service)](/guardrails/gcp-orgpolicy-shared-reservations-owner-projects/)

## Sources

- Define access to models on Vertex AI: https://cloud.google.com/vertex-ai/generative-ai/docs/control-model-access (read 10 Nov 2026)
- How to test it first: https://cloud.google.com/resource-manager/docs/organization-policy/dry-run-policy (read 10 Nov 2026)

Last reviewed 16 Nov 2026. Canonical: https://cloudgovernancetools.com/guardrails/gcp-orgpolicy-vertexai-allowed-models/
