A news publisher's Foundry-hosted assistant drafts satirical opinion pieces that use pointed language about public figures but must never generate content depicting graphic violence. Using the default content safety configuration blocks too many legitimate satire drafts for mild insults while still occasionally allowing violent content through at the default threshold. What should the team do?
Select an answer to reveal the explanation.
Short Explanation
The publisher has two different problems pulling in opposite directions: satire about public figures keeps getting flagged as insulting when it should be allowed, while genuinely violent content is still slipping through. A single shared strictness dial cannot serve both needs at once. The fix is treating each harm category on its own terms, loosening the threshold for pointed or insulting language so legitimate satire gets through, while tightening the threshold specifically for violence so graphic content gets caught more reliably. Turning the filter off entirely solves the satire problem by creating a much bigger one, since violent content would now sail through unchecked. Cranking every category stricter at once fixes nothing, since it makes the satire-blocking problem worse instead of better. And a blocklist built around public figures' names cannot tell an insult from a violent threat from a harmless mention, so it blocks the wrong things for the wrong reasons.
Full Explanation
The correct answer is A. Azure AI Foundry's content safety configuration lets each harm category be tuned independently, so the team can raise the threshold for hate/insult content to let more pointed satirical language through while lowering the threshold for violence so graphic content is blocked more aggressively, matching the two different needs described in the stem. Option B is incorrect because disabling the filter entirely removes protection against the violence category too, the exact risk the publisher must not allow. Option C is incorrect because tightening every category uniformly would make the insult-related blocking worse, not better, continuing to reject legitimate satire drafts. Option D is incorrect because a keyword blocklist keyed on public figures' names would block drafts that mention those figures regardless of whether the content is violent, failing to distinguish the harm type the team actually cares about and blocking harmless satire in the process.