The correct choice is Establish a review process that audits AI generated code for bias and enforces data minimization so sensitive attributes such as gender race or age are not introduced without necessity.

This approach introduces human oversight with clear guardrails. A structured review and audit process helps teams detect biased patterns in prompts and outputs and it ensures code changes follow principles of least data and purpose limitation. You can require pull request reviews and use checklists that call out bias risks and sensitive attributes so engineers justify any use of personal data. You can also combine reviews with automated checks such as linters and security scans that run before merge so the process is both consistent and accountable.

Data minimization reduces the chance that sensitive attributes appear in code or configuration unless there is a clear and documented need. When reviewers enforce this standard during pull requests the team prevents unnecessary collection or propagation of personal data and they reduce downstream risk for users and systems.

Accept Copilot suggestions without edits because they are tuned for speed and accuracy is incorrect because AI generated code can still be biased or insecure and unreviewed acceptance removes essential human oversight and increases risk.

Use Cloud DLP to automatically remove sensitive data from repositories to guarantee ethical AI usage is incorrect because data loss prevention can help with secrets and patterns but it does not audit generated code for bias and it cannot guarantee ethical use. Ethical and responsible use requires policy and human review in addition to tooling.

Configure an automated pipeline that merges any Copilot output without human review because the model is trained to avoid bias is incorrect because models can still produce biased or harmful suggestions and removing human review contradicts responsible engineering practices.

Full AWS Practitioner Certification Question