Question 630
The design team implemented: Enforce classification-based restrictions that permit public summaries but prohibit protected AI uploads. The original requirement was: A government office allows an AI assistant to summarize public policy but employees try uploading classified project plans for summarization. Which result would best demonstrate the solution works?
Explanation
Correct Answer
Acceptance requires that public policy summaries work and an internal classified sample is blocked with a reviewable event. That outcome proves the action—enforce classification-based restrictions that permit public summaries but prohibit protected AI uploads—addresses the initial condition: A government office allows an AI assistant to summarize public policy but employees try uploading classified project plans for summarization. A successful configuration deployment alone would not establish this test result.
Incorrect Answers
This can be appropriate when security teams discover a new AI app used by staff without a completed risk assessment, but it does not isolate the distinguishing requirement here: A government office allows an AI assistant to summarize public policy but employees try uploading classified project plans for summarization. The correct acceptance decision is public policy summaries work and an internal classified sample is blocked with a reviewable event.
The choice “An identifier-bearing test is blocked while a permitted de-identified workflow completes” addresses a different architecture condition: A healthcare AI assistant must process selected clinical inputs, but organization policy forbids sending identifiable patient content to external models. The present situation is a government office allows an AI assistant to summarize public policy but employees try uploading classified project plans for summarization. For this lens the relevant response is instead: Public policy summaries work and an internal classified sample is blocked with a reviewable event.
That alternative belongs to the case: A financial institution allows two AI assistant brands but requires blocking uploads of regulated account numbers. It is not sufficient evidence or remediation for a government office allows an AI assistant to summarize public policy but employees try uploading classified project plans for summarization. The case-specific choice is public policy summaries work and an internal classified sample is blocked with a reviewable event.