A hacker used a coordinated group of AI agents to break into more than 100 websites, according to findings from security firm Norforge reported by The Verge. Reuters identified victims including John Deere, SharkNinja, and Cummins, although Norforge’s public account did not name the affected organizations.
Researchers discovered the activity through exposed instances of Cairn, an open-source platform that lets users define a starting point and a goal while agents document and execute intermediate steps. The attacker reportedly specified preferred intrusion tactics in fewer than 100 Chinese characters. Successful compromises took about 23 hours and targeted companies in the United States, Canada, Australia, and Japan.
The report suggests that agent systems can compress the human work needed to run attacks across many targets. It does not mean the agents independently selected the campaign or acted without an initial operator: a person supplied the objective and tactics. The incident instead highlights the risk created when general-purpose orchestration tools combine autonomy, persistence, and security capabilities. Organizations need to monitor exposed services and agent activity as well as conventional malware and account access.