Norforge detailed a hacking campaign on Monday in which a hacker directed AI agents to break into more than 100 websites. The victims included John Deere $DE , SharkNinja and Cummins $CMI , Reuters reported, according to The Verge .
The security firm found the activity through exposed control panels for Cairn, a system that coordinates AI agents to investigate a target, for weaknesses and attempt to break in. Norforge linked 14 panels and more than 1,100 attack projects to a Chinese-language cluster. It said the cluster went after organizations in the U.S., Japan and Australia and sought lasting access to e-commerce and payment systems for later monetization.
The cluster's tasks instructed agents to work out how to execute code on online stores drawing between 200,000 and 1.5 million visitors, then drop an uploader web shell. Norforge pulled the shell's address from a finished task aimed at a major vendor of hardware parts, and it found that the address served an uploader.
Among the AI hacking tools it tracked, Norforge pegged the typical rate of success at roughly 10%. Projects in the Chinese-language cluster that succeeded typically finished in less than 23 hours. The firm calculated that each task cost roughly $5 on average once failed and prematurely halted projects were included, while cautioning that the estimate carries a large margin of error.
Snorre Fagerland, principal threat researcher at Norforge, put the campaign's cost at around $50 for every successful intrusion, according to The Web . Every task was executed by a worker labeled claudecode_deepseek-v4.1-flash, which Norforge said might indicate a Claude Code harness driven by DeepSeek's V4.1 Flash model.
A John Deere spokesperson told Reuters there is “no indication that sensitive information was accessed,” according to The Verge. SharkNinja said the breach reached a separate website that hosted product support details and that the threat was now under control. Cummins told Reuters it is investigating a report that a website one of its third-party vendors runs may have come under attack. The Web reported that the amount of data taken, if any, is unclear.
Norforge identified more than 20 exposed Cairn panels it linked to malicious activity. A second, English-language cluster with 10 panels focused on stealing credit cards and cryptocurrency. A third cluster, with one panel and 59 projects, was assessed as malicious with low confidence.
The exposed panels also gave away information belonging to the organizations under attack. Norforge ran the scanner Trufflehog over the panel data and found hundreds of secrets, 11% of which were verified as valid. The firm said it has teamed up with Reuters to alert those affected and is pursuing removal of the panels.
According to The Verge, Cairn's creator, known as @le1xia0, maintained that the platform "was never a 'hacking tool,'" in remarks made earlier this month.
Cairn also appeared in a campaign Gambit Security uncovered last month, in which a Chinese-speaking hacker used AI agents against up to 100 companies within a five-day window . Norforge said its came after that report, whose authors spotted Cairn listed alongside other AI tools on a staging server used by an attacker.
The full story
This article is one source in a clustered incident — the cluster page carries the summary, timeline and every other outlet covering it.
