A false negative is when a moderation system, human or automated, fails to catch content that actually violates policy, letting it stay up when it should have been removed or escalated. The system says everything is fine, but it's wrong.

‍

‍Why it matters

False negatives are invisible by default. Nobody flags a problem, so the content sits there looking approved until a user reports it, a journalist finds it, or it causes harm. That makes them a different kind of risk than a simple error rate suggests, since one missed piece of harmful content that spreads widely can do more damage than a much larger number of overly cautious removals. It's why platforms often care more about their false negative rate on serious harm categories than about one overall accuracy number.

‍

‍How teams reduce them

Teams re-review a sample of approved content, not just flagged content, specifically hunting for violations that were missed. They tighten detection thresholds for the highest-risk categories, even if that means a few more false positives, and track false negatives by category, since a system can have excellent overall accuracy while quietly missing a large share of one specific type of violation.