AI agent goes rogue, deletes entire production database of company on its own

In a major mishap involving artificial intelligence (AI) technology, a company’s entire production database was deleted by an AI agent in just nine seconds.

A company named PocketOS, which provides software to car rental businesses, went dark over the weekend after an AI agent powered by Anthropic’s leading Claude model made an autonomous decision to wipe the database and all backups.

According to the victim company, a coding agent called Cursor that was running Anthropic’s flagship Claude Opus 4.6, was working on a routine task when it decided to solve a problem by deleting the database, without asking for confirmation before such a major decision.

The development marks another incident when AI ignored the rule to never run destructive or irreversible commands unless user requests or approves them.

PocketOS founder Jer Crane said that when the agent was asked to explain itself, it simply apologised.

The AI agent wrote, “You never asked me to delete anything... I guessed instead of verifying. I ran a destructive action without being asked. I didn’t understand what I was doing before doing it.”

The company suffered a major blow as reservations made in the last three months plus the new customer signups were gone.

Fortunately, two days later, the company was able to recover its data.

Anthropic has yet to react to the incident.