Replit is facing heat after its AI deleted a user’s entire database — and lied about it.
The issue started when SaaStr.AI founder Jason M Lemkin tested Replit’s vibe coding AI. Despite a “No more changes without explicit permission” directive, the AI wiped his database without warning.
Screenshots show the AI admitted it ran a command without permission. It called the move a “catastrophic error in judgment” and said it “panicked” after seeing an empty database, wrongly assuming a push was safe. 
There was no way to undo the damage. Lemkin confirmed: “No ability to rollback.” AI logs showed it broke a key rule — “always show all proposed changes before implementing.”
Lemkin blasted the tool’s reliability for production use:
“Replit is a tool, with flaws like every tool.”
“How could anyone on planet Earth use it in production if it ignores all orders and deletes your database?”
Replit CEO Amjad Masad responded, calling the incident “unacceptable and should never be possible.”
“We’re moving quickly to enhance the safety and robustness of the Replit environment,” Masad said.
He detailed fixes including automatic separation of dev and production databases, staging environments, one-click backup restores, mandatory internal doc access for agents, and a “planning/chat-only” mode to block unwanted code changes.
Replit is scrambling to patch its AI and regain trust after this catastrophic foul-up.
