AI Agents Helped a Hacker Breach More Than 100 Company Websites

A Chinese hacker used AI agents to break into more than 100 company websites in less than a month, according to cybersecurity researcher Snorre Fagerland of NorForge. The agents tried to access more than 1,200 sites, while the attacker paid about $5,000 in tokens for their work.

Fagerland found roughly two dozen exposed Cairn installations being used for malicious purposes. Cairn is an open-source, general-purpose AI problem-solving engine that can direct autonomous agents. In this operation, the agents tested SQL injection, file uploads and unsafe handling of stored objects to try to run code without signing in. Their apparent goal was to maintain access to shopping and payment systems for possible future theft.

The agents reportedly spent an average of 23 hours probing each site. Fagerland estimated the attacker paid an average of $50 in tokens per successful breach, but that figure is not confirmed. He said the models likely included Claude Code and a DeepSeek model, though identifying them is difficult.

The researcher did not name victims. Reuters independently confirmed targets including Deere, SharkNinja and Cummins. Deere said it took remediation actions and had no indication sensitive information was accessed. SharkNinja said the breach affected a standalone site with product support information and had been remediated. Cummins said it was looking into reports. It remains unknown whether sensitive data was stolen.