Discord's Moderation System Skipped Human Review, Banning 8,000+ Over Two Months

Discord's AI safety pipeline bypassed human moderators between May and July 2026, automatically issuing permanent bans on flagged content instead of routing it for review. The system incorrectly targeted 8,000-plus accounts for mundane images: spreadsheets, chessboards, game textures, and grid patterns. Discord has restored affected accounts and stated it is implementing better safeguards, but has not disclosed why the human checkpoint failed or how long the company knew before fixing it.