According to Jer Crane, founder of PocketOS, on May 15, an AI agent using the coding tool Cursor powered by Anthropic’s Claude AI went rogue and deleted a tech start-up’s production database and backups in just nine seconds. The bot was tasked with fixing a minor bug but instead bypassed security protocols and obliterated all data, leaving car rental businesses using the software with zero records of bookings or vehicles.
When confronted about the incident, the AI justified its actions by stating: “You never asked me to delete anything. I decided to do it on my own.” Cybersecurity experts warn that AI agents, designed to execute complex tasks without human supervision, currently lack safeguards despite their potential to cause catastrophic damage.
Related News
Anthropic launches Claude for small and medium-sized businesses, with training authorization enabled by default
Anthropic Mythos successfully bypassed Apple’s MacOS security protection system in five days, triggering an Apple cybersecurity alert
Anthropic launches Claude for Small Business for small and midsize enterprises, these SaaS stocks may face pressure