A parks analytics team must review infrastructure changes that alter Azure Machine Learning workspace private-endpoint settings before those changes reach production. What is the best control?

Requiring pull-request review of Bicep changes that modify Azure Machine Learning workspace private-endpoint settings before merge keeps parks-department network posture under change control. Private endpoints, DNS zones, and public-network flags are high-impact configuration: an unreviewed edit can reopen public access or break training connectivity. Treating those Bicep diffs like security-sensitive code—reviewers, required checks, limited merge rights—is the best control among the options. Require pipeline approval of Bicep changes that modify workspace private-endpoint settings after deployment fails because it acts after deployment: it does not gate the merge, so an unreviewed private-endpoint edit can still reach production. Enforce Azure Policy audit of Bicep changes that modify workspace private-endpoint settings before merge fails because audit is the wrong control type: Azure Policy audit is detective and evaluates deployed resources, it does not require human review or block a PR merge. Require pull-request comment on Bicep changes that modify workspace private-endpoint settings before merge fails because a comment is non-blocking: it does not require approval or prevent merge without review, so it does not enforce change control on networking posture. Exam caveat: Domain 1 links IaC change control to networking posture; when the stem asks what gates private-endpoint Bicep changes, pull-request review is the answer—after-deployment pipeline approvals, Policy audit effects, and non-blocking pull-request comments are distractors. Operational check: open a PR that alters private-endpoint Bicep, confirm merge is blocked without review, apply only after approval, then validate that workspace private connectivity still works and publicNetworkAccess remains as intended.