[morkalork]

So the premise is that bad actors co-opt the target's hashtags, mass upload ban-able content with those tags, then mass-report that same content and as a casualty the tag gets caught up the next time content moderation models are trained? If the sources aren't just making this up, then it's pretty damning on Facebook's part. Any intern-level data scientist could inspect the training results and see the innocent tags being negatively weighted and flagged.