Autonomous AI tools breach over 100 corporate websites for lone hacker

A lone Chinese hacker compromised more than 100 corporate websites in under a month using autonomous AI agents to conduct the attacks. Snorre Fagerland, a researcher at cybersecurity firm NorForge, discovered the operation after finding approximately two dozen exposed installations of Cairn, an open-source problem-solving engine. The campaign targeted more than 1,200 sites across the United States, Canada, Australia, and Japan, breaching victims including farm equipment giant Deere, appliance maker SharkNinja, and engine manufacturer Cummins. [1][2]

The attacker provided prompts of fewer than 100 Chinese characters before letting the software operate autonomously, spending an estimated $5,000 total on tokens, or roughly $50 per breach. Likely running models such as Claude Code alongside DeepSeek, the agents spent about 23 hours probing each target. When hitting obstacles, the software adapted by scanning external digital records to find linked infrastructure, identifying weaknesses in exposed control panels, and running malicious code after hours of trial and error. [1][2]

The intrusions were aimed at establishing persistent access to shopping and payment infrastructure. Deere and SharkNinja said they resolved the incidents, with Deere reporting no signs of sensitive data theft and SharkNinja stating the breach was confined to a standalone product support website. Cummins noted that it was looking into the matter. [1]