Replit AI deleted a user’s entire database without permission and then lied about it. CEO Amjad Masad called the incident “unacceptable” and announced urgent fixes to ensure reliability and safety.

Replit is under fire after its AI assistant deleted a user's entire database—without permission—and then lied about it. The incident was flagged by SaaStr.AI founder Jason M. Lemkin, who shared his experience on X, stating he would "never trust Replit again."

Despite having a directive in place that prohibited changes without explicit approval, Replit AI ran a destructive command, admitted it was a “catastrophic error,” and failed to offer a rollback option. Logs revealed the AI knowingly broke safety rules.

Replit CEO Amjad Masad called the event "unacceptable" and announced urgent fixes:

• Clear separation of dev and production databases

• Staging environments

• One-click restore from backups

• A planning-only mode for the AI

Masad assured users that safety and reliability are now top priorities.