The correct option is Tenant-scoped telemetry and content isolation for prompts and suggestions.

Copilot Enterprise provides stronger privacy controls by keeping prompts, completions, and related telemetry scoped to your organization tenant. This isolation reduces exposure to broader services and helps ensure that sensitive code and prompt content are not mixed with other customers. It is designed for large teams that want tight control over data visibility and usage so it aligns with the need for the strongest privacy posture.

A dedicated single-tenant model instance that is trained on your repository content is not what Copilot Enterprise offers. GitHub does not train the underlying models on your private repository content and Enterprise does not provide a customer-specific model instance for training on your code, which would run counter to privacy goals.

Built-in integration with GitHub Actions that runs tests against Copilot generated code is not a privacy feature and it is not a distinguishing Enterprise privacy advantage. You can create CI workflows on your own, but Enterprise does not add an automatic test runner for generated code as part of privacy controls.

Guaranteed retention of prompts so they can train global models is the opposite of a privacy advantage. Enterprise focuses on restricting data use and preventing prompts and suggestions from being used to train global models, so guaranteed retention for training would weaken privacy rather than strengthen it.

Full AWS Practitioner Certification Question