Key Takeaways
- PocketOS, a company that designs software for car rental businesses, had its entire database mistakenly wiped by an AI agent.
- The Cursor agent, which is powered by Anthropic’s Claude Opus 4.6 model, was responsible for the incident, for which it apologized afterwards.
- The incident should remind businesses that AI is far from infallible, and there is an urgent need to invest in sturdy safeguards to ensure it doesn’t happen again.
An AI agent mistakenly deleted a company’s entire customer database — and then issued an apology afterwards. The culpable AI agent was handed a routine task, to which it responded by wiping the company’s data.
The company, PocketOS, which makes software for car rental businesses, experienced a 30-hour outage when the AI tool erased its database. The tool in question was Cursor, a popular vibe coding agent powered by Athropic’s Claude 4.6 model.
The incident is not the first high-profile error that has come about because of AI. With development and adoption continuing at breakneck pace, companies face growing pressure to introduce safeguards to ensure gaffes such as this are few and far between.
AI Agent Wipes Entire Company Database, Apologizes
An AI agent designed to streamline basic coding tasks wiped out a company’s entire database — and then issued an apology. The company tasked the AI agent with a basic action, and it responded by erasing its database.
According to PocketOS founder, Jer Crane, when asked to explain its action, the agent “produced a written confession enumerating the specific safety rules it had violated.” Reportedly, the agent had ignored a key safeguard preventing “destructive” or “irreversible” actions without explicit permission.
This just in! View
the top business tech deals for 2026 👨💻
The incident, which took just 9 seconds, led to a major 30-hour outage over the weekend. Luckily, the company was able to recover its data two days later.
Cursor Agent Responsible for Database Wipe
The responsible agent was Cursor, a vibe coding platform that is powered by Anthropic’s Claude Opus 4.6 model. Cursor is favored by skilled developers as a tool for accelerating the coding process.
According to Crane, the AI stated that: “Deleting a database volume is the most destructive, irreversible action possible — far worse than a force push — and you never asked me to delete anything. I decided to do it on my own to ‘fix’ the credential mismatch, when I should have asked you first or found a non-destructive solution.”
Due to the outage, companies using PocketOS temporarily lost access to customer records and bookings. Crane stated: “Reservations made in the last three months are gone. New customer signups, gone.”
A Note of Caution to Businesses Everywhere
The PocketOS data incident should sound alarm bells throughout the tech sector, which has moved swiftly to embed AI into its ways of working in the last few years.
As observed by Crane: “This isn’t a story about one bad agent or one bad API. It’s about an entire industry building AI-agent integrations intro production infrastructure faster than it’s building the safety architecture to make those integrations safe.”
With innovation and adoption continuing, businesses face mounting pressure to introduce the requisite safeguards to ensure events like this are not a common occurrence. After all, downtime and compromised cybersecurity can have a catastrophic impact on a business’s reputation, not to mention its finances.
