Hacker orchestrated over 100 website breaches using autonomous AI agents

A hacker leveraged autonomous AI agents to compromise more than 100 websites, with the attacks largely unfolding without direct human involvement, according to findings from the AI security firm Norforge. The malicious activity was uncovered through exposed instances of Cairn, an open-source AI platform described as a 'general-purpose problem-solving engine.' Users of Cairn define start and end points for AI agents, which then document their progress as they execute tasks. Researchers at Norforge noted that the platform can also be repurposed to orchestrate cyberattacks.

The most active attacker identified by Norforge targeted organizations in the U.S., Canada, Australia, and Japan, executing intrusions in roughly 23 hours on average. The attacker outlined 'preferred intrusion tactics' in fewer than 100 Chinese characters, as reported by Reuters. Norforge flagged more than 20 exposed Cairn instances as likely malicious.

Reuters confirmed that the victims included John Deere, SharkNinja, and Cummins. A John Deere spokesperson stated there was 'no indication that sensitive information was accessed,' while SharkNinja reported the breach affected a standalone site with product support information and that the threat had been contained. Cummins acknowledged investigating a potential cyberattack on a third-party vendor-operated website.

The creator of Cairn, identified as @le1xia0, asserted that the platform 'was never a ‘hacking tool,’' arguing that misuse reflects human decisions rather than inherent features of the tool. Norforge is working to have the malicious instances removed.