AI agents broke into more than 100 company sites with little human direction
A security researcher at NorForge, Snorre Fagerland, says he watched a Chinese-speaking hacker use AI agents to break into websites belonging to more than 100 companies, after finding more than 20 exposed control panels of an open-source agent platform called Cairn. According to his report, the hacker gave short instructions and left the agents to probe and attack on their own. Reuters, which saw the report before publication and identified victims including Deere, SharkNinja and Cummins, reports that the agents attempted more than 1,200 sites, that a successful break-in took around 23 hours, and that the campaign used a mix of DeepSeek and Anthropic's Claude Code. Fagerland estimates it cost roughly $5,000 in total, or about $50 per successful hack; he says these figures are estimates. Deere said it had taken remediation steps and had no sign that sensitive data was accessed.
Why it matters: This is early evidence that cheap AI agents can run real intrusions with little skill or money, so treat your public-facing sites and vendors' sites as higher risk and make sure patching and monitoring are current.
NorForge ↗AI-assisted, reviewed by Sorted AI
▲ 0
0 comments