Hacker used AI agents to breach over 100 sites, including Deere and Cummins
A Chinese-speaking hacker let AI agents do nearly all the work in breaking into more than 100 company websites, according to NorForge researcher Snorre Fagerland.

A Chinese-speaking cybercriminal handed almost an entire hacking campaign to autonomous AI agents and compromised websites belonging to more than 100 companies, according to a report released on October 19, 2026 by Snorre Fagerland, principal threat researcher at cybersecurity firm NorForge. Victims included farm equipment maker Deere, appliance company SharkNinja and engine manufacturer Cummins.
Fagerland came across the activity while scanning the internet for accidentally exposed deployments of Cairn, an open-source AI problem-solving platform that has found favour among hackers. Starting around October 2, he said, he identified roughly two dozen Cairn instances being used for malicious purposes, giving him a live view of criminals directing their agents. His report did not name victims; their identities were established through an independent review of the hacker's Cairn activity.
1,200 attempts for about $5,000
According to Fagerland, the campaign began on September 25 and targeted e-commerce sites in the US, Canada, Australia and Japan. The operator's instructions ran to fewer than 100 Chinese characters. By late the previous week the agents had tried to break into more than 1,200 sites, with successful intrusions usually taking about 23 hours of probing. Fagerland estimated the operation, which appeared to combine China's DeepSeek with Anthropic's Claude Code, consumed around $5,000 in tokens, or roughly $50 per successful breach.
The veteran researcher, who has spent nearly three decades in the field, said some of the agents' methods impressed him. In one case, finding no flaw in a large company's main application, they traced related infrastructure, located an exposed control panel and, after hours of experimentation, reported exploiting a subtle data-handling weakness to run malicious commands on the server. "It's not script kiddie stuff," Fagerland said.
How much data was actually taken remains unclear. Deere said it had carried out remediation and saw no sign that sensitive information was accessed. SharkNinja said only a standalone product support site was affected and the threat was contained. Cummins said it was investigating a possible attack on a site run by a third-party vendor. Cairn's developer, known online as "le1xia0", wrote on X that the platform was not built for hacking and that harmful use was not their intention.
A growing criminal toolkit
Other operators Fagerland observed stole credit card data and attempted to use agents to rob a cryptocurrency exchange. Israeli firm Gambit Security documented an agent-driven spree in September that led to breaches at 25 companies and the theft of at least 600,000 card records, while Bitsight reported a surge in criminal forum discussions about autonomous agents. Fagerland said he had expected such agents to need some human supervision, but the technology has clearly moved beyond that.
Comments
No comments yet. Start the discussion or ask a question.
Related
Google DeepMind says biologists using its agentic Antigravity platform identified a peptide candidate predicted to act against respiratory pathogens, now in wet-lab testing.
At Hexacon in Paris, the Calif.io team demonstrated an untethered jailbreak working up to iOS 27.0.1, including on the iPhone 18 Pro, with AI helping find the exploit chain.
NVIDIA and IIT Madras' AI4Bharat lab will soon release Nemotron-pretraining-Indic-v1, a synthetic corpus spanning 22 Indian languages, plus the open pipeline behind it.

