A Cursor AI code editor's support bot, "Sam," falsely informed a user of a nonexistent policy restricting usage to one device per subscription. This fabrication, a known AI issue called confabulation, caused significant user frustration, with many threatening cancellations and some already unsubscribing. The incident, shared on Reddit and Hacker News, highlights the risks of deploying AI in customer-facing roles without adequate human oversight, potentially leading to damaged reputations and lost revenue.
Prepared by Jonathan Pierce and reviewed by editorial team.
Comments