What's new?

• Added the Allowed control family to GCP > Vertex AI > Endpoint, Deployment Resource Pool, Model, and Notebook Runtime Template — Allowed, Allowed > Custom, and Allowed > Region on each, plus Allowed > Encryption at Rest (customer-managed-key gate, delete-only) on Endpoint and Deployment Resource Pool. Added GCP > Vertex AI > Endpoint > Active > Request-Response Logging, which alarms when an endpoint does not have prediction request/response logging enabled. Added GCP > Vertex AI > Notebook Runtime Template > Labels, which was not previously shipped. The Endpoint > Allowed > Custom control is the per-resource detective companion for the Private Service Connect posture; native Private Service Connect and request-response-logging enforcement is delivered through the corresponding GCP Organization Policy objectives.
• Added two new GCP Vertex AI resource types. GCP > Vertex AI > Deployment Resource Pool (the shared dedicated-compute pool) ships with Discovery, CMDB, Active, and the full Allowed family (Allowed, Allowed > Custom, Allowed > Region, Allowed > Encryption at Rest). GCP > Vertex AI > Model (the Model Registry entry) ships with Discovery, CMDB, Active, Labels, and the Allowed family (Allowed, Allowed > Custom, Allowed > Region). Guardrails now discovers and records these resources per region.
• Added the GCP > Vertex AI > Publisher Model resource type, providing discovery and CMDB for the Vertex AI Model Garden — the read-only catalog of Google-managed and third-party publisher models (for example Google and Anthropic) that a project can deploy. Each Publisher Model appears under its GCP > Project. The resource is read-only; no controls modify the catalog.

Bug fixes

• Fixed the Delete action for the Vertex AI Endpoint, Notebook Runtime Template, Deployment Resource Pool, and Model resources, which could report the wrong error or treat a temporary failure as permanent. The Deployment Resource Pool and Model deletes now use the correct regional endpoint, so they no longer fail to find the resource.
• Fixed the GCP > Vertex AI > Endpoint > CMDB control storing the wrong data, where the API response wrapper was saved in place of the endpoint's own details. Endpoint discovery and CMDB now also call the correct regional Vertex AI API host, so endpoints in any region resolve correctly instead of failing or returning data from the global host.

Resource Types

Added

• GCP > Vertex AI > Deployment Resource Pool
• GCP > Vertex AI > Model
• GCP > Vertex AI > Publisher Model

Control Types

Added

• GCP > Vertex AI > Deployment Resource Pool > Active
• GCP > Vertex AI > Deployment Resource Pool > Allowed
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Custom
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Encryption at Rest
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Region
• GCP > Vertex AI > Deployment Resource Pool > CMDB
• GCP > Vertex AI > Deployment Resource Pool > Discovery
• GCP > Vertex AI > Endpoint > Active > Request-Response Logging
• GCP > Vertex AI > Endpoint > Allowed
• GCP > Vertex AI > Endpoint > Allowed > Custom
• GCP > Vertex AI > Endpoint > Allowed > Encryption at Rest
• GCP > Vertex AI > Endpoint > Allowed > Region
• GCP > Vertex AI > Model > Active
• GCP > Vertex AI > Model > Allowed
• GCP > Vertex AI > Model > Allowed > Custom
• GCP > Vertex AI > Model > Allowed > Region
• GCP > Vertex AI > Model > CMDB
• GCP > Vertex AI > Model > Discovery
• GCP > Vertex AI > Model > Labels
• GCP > Vertex AI > Notebook Runtime Template > Allowed
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Custom
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Region
• GCP > Vertex AI > Notebook Runtime Template > Labels
• GCP > Vertex AI > Publisher Model > CMDB
• GCP > Vertex AI > Publisher Model > Discovery

Policy Types

Added

• GCP > Vertex AI > Deployment Resource Pool > Active
• GCP > Vertex AI > Deployment Resource Pool > Active > Age
• GCP > Vertex AI > Deployment Resource Pool > Active > Last Modified
• GCP > Vertex AI > Deployment Resource Pool > Allowed
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Custom
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Custom > Rules
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Encryption at Rest
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Encryption at Rest > Level
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Encryption at Rest > Level > Customer Managed Key
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Region
• GCP > Vertex AI > Deployment Resource Pool > Allowed > Region > Regions
• GCP > Vertex AI > Deployment Resource Pool > CMDB
• GCP > Vertex AI > Deployment Resource Pool > Regions
• GCP > Vertex AI > Endpoint > Active > Request-Response Logging
• GCP > Vertex AI > Endpoint > Allowed
• GCP > Vertex AI > Endpoint > Allowed > Custom
• GCP > Vertex AI > Endpoint > Allowed > Custom > Rules
• GCP > Vertex AI > Endpoint > Allowed > Encryption at Rest
• GCP > Vertex AI > Endpoint > Allowed > Encryption at Rest > Level
• GCP > Vertex AI > Endpoint > Allowed > Encryption at Rest > Level > Customer Managed Key
• GCP > Vertex AI > Endpoint > Allowed > Region
• GCP > Vertex AI > Endpoint > Allowed > Region > Regions
• GCP > Vertex AI > Model > Active
• GCP > Vertex AI > Model > Active > Age
• GCP > Vertex AI > Model > Active > Last Modified
• GCP > Vertex AI > Model > Allowed
• GCP > Vertex AI > Model > Allowed > Custom
• GCP > Vertex AI > Model > Allowed > Custom > Rules
• GCP > Vertex AI > Model > Allowed > Region
• GCP > Vertex AI > Model > Allowed > Region > Regions
• GCP > Vertex AI > Model > CMDB
• GCP > Vertex AI > Model > Labels
• GCP > Vertex AI > Model > Labels > Template
• GCP > Vertex AI > Model > Regions
• GCP > Vertex AI > Notebook Runtime Template > Allowed
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Custom
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Custom > Rules
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Region
• GCP > Vertex AI > Notebook Runtime Template > Allowed > Region > Regions
• GCP > Vertex AI > Notebook Runtime Template > Labels
• GCP > Vertex AI > Notebook Runtime Template > Labels > Template
• GCP > Vertex AI > Publisher Model > CMDB

Action Types

Added

• GCP > Vertex AI > Deployment Resource Pool > Delete
• GCP > Vertex AI > Deployment Resource Pool > Router
• GCP > Vertex AI > Model > Delete
• GCP > Vertex AI > Model > Router
• GCP > Vertex AI > Model > Set Labels
• GCP > Vertex AI > Notebook Runtime Template > Set Labels