Vertex AI managed models
Google Cloud says it will not use customer data to train or fine-tune managed AI models on Vertex AI without prior permission or instruction.
The Google Cloud training restriction applies to managed models on Vertex AI, including generally available and pre-general-availability models.
No opt-out is required for the stated training restriction. Teams must still configure storage, caching, logging, location, and optional tuning features.
Customer prompts and grounded data remain subject to the chosen Google Cloud services, project controls, regions, and data-processing terms.
Google documents zero-data-retention considerations separately; caching, abuse monitoring, grounding, evaluation, and other features can create retention paths that require review.
The cited data-governance page focuses on training restrictions and service data handling rather than one universal human-review statement.
Foundational reference pages for teams adopting, deploying, and operating production AI with inspectable controls.