Discord’s AI moderation system went rogue, falsely banning over 8,000 users since May after flagging harmless images—spreadsheets, chessboards, game textures, and transparent backgrounds—as illegal content. The bug, which caused automatic bans instead of human review, was finally fixed over the weekend, with accounts being restored. This mess highlights the steep challenges of AI-powered moderation. When algorithms rely on similarity matching to catch harmful material, they can trip over innocent patterns like grids, which users on X and Reddit say are being over-flagged due to past misuse. Discord admitted the flaw came from a bug that bypassed their usual human review step. Affected users are furious, stressing that permanent bans can wreck livelihoods and social ties. Why it matters: This incident underscores a critical blind spot in AI moderation—false positives can cause real damage, and platforms need better safeguards before deploying automated systems at scale.