Discord's AI-powered safety systems mistakenly banned approximately 8,200 accounts since May 2026, due to a bug that flagged grid-like images as child sexual abuse material. The issue affected a wide range of content, from spreadsheets and chessboards to in-game menus like that in Minecraft.

Normally, Discord's AI flags are reviewed by a human Trust & Safety team member before action is taken. However, a bug caused these accounts to be banned immediately instead of having uploads paused for review. The same bug also prevented automatic reinstatement once human reviewers cleared the accounts, leaving bans in place for up to two months.

Discord acknowledged the error and has begun reinstating affected accounts. This incident exposes a significant flaw in automated safety systems and raises questions about the reliability of AI moderation without human oversight. The company stated it is developing better safeguards to prevent similar occurrences.