The correct answer is Your editor sends the prompt to a GitHub proxy for initial policy checks then a Copilot model on remote servers generates output and additional filters run before the IDE shows the suggestions.

This reflects how Copilot processes your request from the IDE to GitHub infrastructure that performs policy and security checks and routing before a hosted model generates code. Copilot applies content and safety filtering before and after generation to enforce policies and quality and the editor then renders the allowed suggestions.

Your input is uploaded and stored on GitHub servers so the model can retrieve it during generation to preserve continuity of suggestions is incorrect because prompts are sent for processing but are not persistently stored to maintain continuity. Continuity comes from the current editor context included in each request and GitHub documents limited retention and no training on your private code by default.

The Copilot extension sends your code to Vertex AI where a code generation model runs and returns the response straight to your editor is incorrect because Copilot does not use Vertex AI to serve completions. Requests flow through GitHub services and models hosted by GitHub and Microsoft and they pass through policy checks and filters before returning to the IDE.

The Copilot model runs locally on your machine and your inputs are saved to improve future model versions is incorrect because the models run in the cloud while the IDE extension is local. Inputs are not saved to train models without enterprise controls or opt in and local execution is not how Copilot delivers suggestions.

Full AWS Practitioner Certification Question