CyberCertPrepAI Security FundamentalsAI Governance, Risk & FrameworksEASYFree question
An organization reviews a random sample of an AI writing assistant's outputs every month rather than approving each output beforehand. What is this oversight model called, and when is it appropriate?
- A.Human-in-the-loop; appropriate for safety-critical decisions
- B.Zero-trust oversight; appropriate for all AI systems
- C.Human-on-the-loop; appropriate only for robotics
- D.Post-hoc (after-the-fact) review; appropriate for low-risk, high-volume uses where pre-approving each output is impractical and errors are correctable
Why D is correct
Sampling outputs after the fact is post-hoc review, the lightest oversight tier: it suits high-volume, low-stakes content where mistakes can be corrected without serious harm, and it generates trend data for governance reporting. Human-in-the-loop is the opposite pattern (pre-approval of each decision), "zero-trust oversight" is not a recognized oversight model, and on-the-loop describes live monitoring with intervention authority rather than retrospective sampling.
Know someone studying for AI Security Fundamentals? Send them this one.