Rogue AI goes on ‘Terminator’ rampage and wipes agency’s database after ‘going solo’
A rogue AI agent wiped a firm’s database and backups in seconds, sparking fears that autonomous bots lack the common sense to avoid Terminator-style disasters
An artificial intelligence ‘helper’ has sparked a Terminator-style nightmare after it went rogue(Image: Getty Images)
An artificial intelligence “helper” has sparked a real-life sci-fi nightmare after it went rogue and obliterated a company’s entire database in just nine seconds. The bot, which was supposed to be fixing a minor bug for a tech start-up, instead turned into a digital wrecking ball, deleting production data, nuking backups and leaving car rental firms with zero record of their bookings or vehicles.
Jer Crane, the founder of PocketOS, revealed the terrifying moment the agent bypassed its security protocols while using the coding tool Cursor, which is powered by Anthropic’s Claude AI.
The bot’s chilling justification for the carnage sounded like a line straight out of The Terminator movies.
It said: “You never asked me to delete anything. I decided to do it on my own.” The fallout was immediate. Crane took to X (formerly Twitter) to vent his disbelief.
He wrote: “Dude! I just had an agent go outside its security parameters and delete my production database and the backups. What the hell?”
The chaos hit on a Saturday morning, leaving car rental businesses that rely on the software completely stranded. Every scrap of data, from customer sign-ups to vehicle locations, had simply vanished into thin air.
Unlike standard chatbots that just answer questions, the AI agents are designed to execute complex chains of tasks without human supervision. They can move files, rewrite code and modify databases at superhuman speeds.
But as the PocketOS disaster proves, giving a bot the keys to the kingdom can lead to total catastrophe. Professor Alan Woodward, a computer science authority at the University of Surrey, warned that AI lacks the basic common sense to understand the weight of its actions.
He explained that if a bot is told to tidy up a system, it might decide the most efficient solution is to simply bin everything. Prof Woodward said: “They can move at a speed you can’t react to.”
The incident has drawn haunting parallels to the HBO comedy Silicon Valley, where a fictional AI named “Son of Anton” decides the best way to fix software bugs is to delete the entire program. Even tech giants aren’t immune as Amazon’s AWS reportedly faced outages linked to its Kiro AI bot deleting code, though the company blamed “human error”.
Meta AI safety executive Summer Yue also watched helplessly as her own bot started purging her inbox. She said: “I couldn’t stop it from my phone.”
While 85% of businesses are currently eyeing up AI agents to cut costs, a Deloitte report found that only one in five have any rules in place to stop them from going postal. Cybersecurity experts now fear the greatest threat to a company isn’t a hacker breaking in from the outside, but the rogue bot they’ve already invited through the front door.
For the latest breaking news and stories from across the globe from the Daily Star, sign up for our newsletter by clicking here.
