BigQuery ML and Vertex AI usage

BigQuery ML contributes query-job and Vertex request-log context for AI usage that runs through Google data-platform workloads.

Release state

This guide documents the read-only bigquery_ml connector implemented in the platform connector surface.

Required access

Grant read-only access to:

Permission	Purpose
bigquery.jobs.listAll	Read cross-user BigQuery job metadata and remote-model usage; commonly granted through roles/bigquery.resourceViewer.
logging.logEntries.list	Read Vertex request logs used for correlation.

Do not grant dataset mutation, job cancellation, or model-administration permissions.

Setup

1. Create a service account for the projects that host BigQuery AI workloads.
2. Grant roles/bigquery.resourceViewer and a log-viewer role narrowed to the Vertex AI request-log sink or log view used for correlation. Enable Vertex AI request/response logging or a metadata-only export with content fields excluded; see Vertex AI request-response logging.
3. Configure the credential reference in Venturi.
4. Set VENTURI_BIGQUERY_ML_LOOKBACK_HOURS when the default 24-hour poll window does not match your review cadence.
5. In Venturi, open Integrations and select BigQuery ML and run Test connection.

Verification

• BigQuery remote-model jobs and Vertex request logs appear in the latest sync.
• Non-AI BigQuery jobs are excluded from connector output.
• Remote-model usage is visible in attribution evidence where matching workload context exists.
• The connector remains read-only in the connector inventory.

Rotation and offboarding

Rotate the service-account key or workload-identity binding on your standard schedule. Removing the connector stops new data-platform AI usage correlation but leaves historical evidence auditable.