A customer building a scalable implementation is considering letting end users flag unsafe generative AI outputs as the only safety check. Why is this insufficient on its own?
Select an answer to reveal the explanation.
Short Explanation
You can’t rely on users as the only safety check because the harmful output has already reached them by the time they flag it. You need safety evaluation before or during deployment, not just after-the-fact reporting.
Full Explanation
The correct answer is the option stating that user reporting relies entirely on users noticing and reporting harm after it has already reached them, rather than evaluating outputs for safety before or during deployment. User reporting is reactive: it only catches problems already shown to someone who noticed and chose to report them, missing everything else. The option claiming users are legally prohibited from reporting generative content is incorrect because no such general prohibition exists. The option saying user flags are always inaccurate and should be disregarded entirely is incorrect because flags can be useful; the issue is relying on them exclusively. The option saying user-reported issues cannot be logged or tracked is incorrect because they can typically be logged and tracked. For the exam, treat user reporting as a complement to systematic evaluation, never a substitute for it.