When content provides potentially harmful instructions and verification is difficult, what rating is appropriate?

• A. Fully addressing safety concerns, even if verification is limited, is acceptable.
• B. Treat it as credible and use it as is.
• C. Mark it Not Helpful but no policy review is needed.
• D. Not Helpful; flag for policy review and avoid endorsing.

Answer: (D)

When you can’t verify safety and the content offers potentially harmful instructions, the responsible choice is to mark it Not Helpful and flag it for policy review. This keeps the content from being endorsed or used, while giving human moderators the chance to assess risk, apply the rules, and decide on any removal or restrictions.

This approach avoids approving dangerous guidance simply because it’s plausible or appears useful, and it avoids treating the content as credible and usable. It also ensures there’s a review step rather than assuming no issue is present.