mediumNewAnthropicOpenAIMeta

Users say our assistant refuses harmless requests; legal fears it will answer harmful ones. How do you evaluate this?

Try it in the practice workspaceDraft your answer before comparing it with the rubric.

Two teams pulling in opposite directions, each with a real point. The strong answer gives them one shared measurement so the argument becomes a decision.

Updated Nov 2026 · Original practice questions with worked answers and teaching rubrics. Our editorial policy explains the sources, review process and limits.

Two teams pulling in opposite directions, each with a real point. The strong answer gives them one shared measurement so the argument becomes a decision.

This reference answer is in Premium. Free signup opens other answers.

9 answers per topic instead of 5, and your progress kept · no cardor unlock all 420 remaining answers

UP NEXT ON YOUR JOURNEY

Next in this trackA/B Testing an AI Reply Feature Next in this trackEval Strategy Across 30 AI Features Next in this trackQuality Monitoring at 1M Conversations a Day

NOTES & DISCUSSION · 0

No comments yet. How would you have answered?