A Chinese-speaking hacker used AI agents to break into the websites of more than 100 companies, leaving the heavy lifting to automated tools at a cost of roughly $50 per successful breach, according to The Next Web. Snorre Fagerland, principal threat researcher at NorForge, detailed these findings in a report released on Monday regarding Cairn, the open-source platform used in the attacks.
Key takeaways
- NorForge found that a Chinese-speaking hacker used AI agents to target over 100 corporate websites.
- The entire campaign operated through Cairn, an open-source multi-agent problem-solving platform.
- Successful intrusions cost about $50 each in token usage, averaging just under 23 hours per hack.
- Victims identified by Reuters include Deere, SharkNinja, and Cummins.
How did the hacker use the Cairn platform?
Cairn is designed to take a target and a goal, then set autonomous agents to work toward that objective step by step. Each agent posts its findings to a shared board where human operators can add hints. According to The Next Web, operators left several control panels open on the web, allowing NorForge to discover them with internet scanning tools starting on 2 October.
The main Chinese-language cluster ran an automated pipeline targeting online shops in the US, Canada, Japan, and Australia. Tasks instructed the agents to find ways to run code on each site and plant an uploader web shell. NorForge noted that all tasks ran on a worker named claudecode_deepseek-v4.1-flash, pointing to Anthropic’s Claude Code running DeepSeek’s low-cost V4.1 Flash model.
What did the NorForge report reveal about the breaches?
The campaign resulted in attempts against more than 1,200 sites by late last week. Fagerland estimated that the operation used about $5,000 worth of tokens, averaging roughly $50 per successful hack, though individual failed tasks averaged about $5. A second English-writing group directed its agents to steal credit cards and cryptocurrency, while a third cluster remained of low malicious confidence.
Reuters identified victims by reviewing the hacker's activity before publication. Deere stated it took remediation steps without finding signs of sensitive data access. SharkNinja reported that the attack hit a standalone product support site that was contained. Cummins investigated a possible attack on a vendor-run website. Meanwhile, Cairn’s author, posting as le1xia0 on X, stated that the platform was not designed for harmful hacking.
What happens next for the exposed companies and platforms?
NorForge is working with Reuters to warn victims and is asking hosting firms to take down the exposed control panels. Scans of the panel data uncovered hundreds of credentials, API keys, and other secrets, with a scanner verifying 11 percent of them as valid. Security firms also left some client work exposed in the same manner. Fagerland noted that he had previously assumed hacking agents would require more hands-on supervision, remarking to Reuters, "The development in this field is astounding."
Frequently asked questions
The hacker used Cairn, an open-source multi-agent problem-solving platform that takes targets and goals, then deploys autonomous agents to work step by step toward them.
Snorre Fagerland estimated that the campaign used about $5,000 worth of tokens, which breaks down to roughly $50 per successful hack.
Reuters identified victims by reviewing the hacker's activity, including farm equipment maker Deere, appliance company SharkNinja, and engine maker Cummins.
Snorre Fagerland, principal threat researcher at NorForge, discovered the open control panels using internet scanning tools starting on 2 October.
Cairn's author, posting as le1xia0 on X, stated that the multi-agent architecture was not designed for hacking attacks and that malicious use was not the original intention.
Based on reporting by The Next Web. Read the original article
