Microsoft Copilot Studio
Microsoft says Copilot Studio customer prompts, grounding data, and responses are not used to train or improve Azure OpenAI foundation models.
Copilot Studio prompt-builder customer data is not used to train or improve the foundation models that provide the service.
No opt-out is required for the stated training baseline. Makers and administrators still choose knowledge sources, actions, connectors, regions, and logging.
Copilot Studio can ground prompts in connected business data and invoke workflows, so permissions and connector boundaries remain material even without model training.
Microsoft documents temporary secured conversation storage for operations; feature and environment choices can also create Dataverse, telemetry, or connected-system records.
Microsoft recommends meaningful human oversight for generated content; the reviewed FAQ focuses on responsible operation rather than one support-review disclosure.
Foundational reference pages for teams adopting, deploying, and operating production AI with inspectable controls.