Discord has confirmed that a bug in its AI-powered moderation system mistakenly banned more than 8,000 users over the past two months after harmless images were incorrectly identified as illegal content.
The issue, which affected accounts since May, caused images such as spreadsheets, chessboards, game textures, and even plain white or gray transparent backgrounds to be flagged by the platform’s automated detection system. Discord said another 200 users were mistakenly banned over the weekend before engineers identified and fixed the bug. The company is now restoring all affected accounts.
According to Discord, its moderation system compares uploaded content against databases of known harmful material to detect illegal images. While this similarity-matching technology is designed to improve safety, it can occasionally produce false positives. The company explained that flagged content is normally reviewed by a member of its Trust & Safety team before enforcement action is taken, but a software bug caused some affected accounts to be banned immediately instead.
Discord said it is adding additional safeguards to prevent similar incidents from happening again.
Following the bans, users on X and Reddit reported being permanently suspended after uploading images containing simple square grid patterns. Some speculated that the moderation system had become overly sensitive to grid-like images because similar patterns have previously been used to disguise or evade detection of illegal content.
The mistaken bans sparked frustration among affected users, particularly those who depend on Discord for work, game development, online communities, or staying connected with friends. Some said losing access to their accounts without warning had a significant personal and professional impact.
The incident adds to growing concerns about the reliability of AI-assisted content moderation as online platforms increasingly depend on automated systems to detect harmful material at scale.
Discord is not the only company to face criticism over AI moderation. Last year, users of Instagram and Meta’s Facebook Groups reported widespread unexplained account suspensions that many suspected were caused by automated moderation, although Meta never confirmed whether AI errors were responsible. More recently, Meta’s Oversight Board has called for greater transparency around moderation decisions. Users of Tumblr also reported unexplained mass account suspensions last year, raising similar concerns about automated enforcement systems.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
