Hacker used AI agents to breach 100+ sites — including John Deere — at $50 each
What's the deal? A single hacker used a swarm of AI agents to break into more than 100 websites, including ones linked to John Deere, SharkNinjaDealroom has a profile for this one. Try Dealroom →, and CumminsDealroom has a profile for this one. Try Dealroom →, according to security firm NorForgeDealroom has a profile for this one. Try Dealroom →. The agents carried out the attacks with little involvement from the human, who largely set the goals and tactics.
How it worked: Researcher Snorre Fagerland found the campaign through exposed instances of Cairn, an open-source AI agent platform. The Chinese-speaking hacker gave the agents fewer than 100 characters of instructions, then let them probe more than 1,200 e-commerce sites on their own; successful intrusions took about 23 hours of autonomous work, apparently using a mix of DeepSeek and Claude Code.
Why now? The campaign stands out for scale and cost rather than any novel vulnerability. The whole operation burned roughly $5,000 in tokens — about $50 per successful breach. "It's not script kiddie stuff," Fagerland said.
What could go wrong? The economics suggest attackers can now chase many targets cheaply, automating reconnaissance and intrusion work that once needed skilled humans. Deere said it has no indication sensitive information was accessed; the report points to a growing operational threat, not an established major data breach.
The signal: A single actor with under 100 characters of instructions compromised over 100 companies' sites — a concrete data point for how cheap and autonomous AI-driven hacking has become.
