A Chinese-speaking cybercriminal broke into the websites of more than 100 companies in a few weeks while handing nearly all of the hands-on work to artificial intelligence agents, Reuters reported Monday in an exclusive, citing a security researcher’s findings and its own review of the activity. Deere, SharkNinja and Cummins were among the companies affected. The case suggests the cost and skill needed to attack companies at scale are falling fast.
What Happened
The findings come from Snorre Fagerland, principal threat researcher at the cybersecurity company NorForge, who shared a report with Reuters before publication. NorForge has not published the report, so the details below are the researcher’s account as Reuters reported it.
An AI agent is software built on a large language model that is given a goal and then plans steps, runs tools and adjusts, with little human input. According to Fagerland, the hacker gave the agents brief instructions, fewer than 100 Chinese characters describing preferred intrusion tactics, and the agents then probed targets on their own for hours at a time.
Starting Sept. 25, the agents went after e-commerce websites in the U.S., Canada, Australia and Japan. The agents tried more than 1,200 sites and got into more than 100 companies. A successful break-in typically took about 23 hours of probing. Fagerland estimated the hacker spent roughly $5,000 on AI usage fees in total, or about $50 per successful intrusion. He said the campaign appeared to use a mix of the Chinese model DeepSeek and Anthropic’s coding agent Claude Code, run through Cairn, an open-source AI problem-solving platform.
In one case Reuters described, the agents found no weakness in a large company’s main public application, so they searched public records for related systems, found an exposed control panel and, after hours of trial and error, got the server to run commands of their choosing.
Fagerland came across the activity while scanning the internet for exposed Cairn setups beginning around Oct. 2. He found about two dozen being used for malicious purposes; other operators were stealing payment card data or trying to rob a cryptocurrency exchange. The researcher’s report did not name victims. Reuters identified them by reviewing the hacker’s activity on the platform.
Deere told Reuters it had taken remediation steps and had no indication that sensitive information was accessed. SharkNinja said the intrusion hit a standalone site holding only product support information and that the threat was contained. Cummins said it was investigating a possible attack on a website run by one of its third-party vendors. Two smaller businesses confirmed being compromised, and Reuters said it could not determine how much data, if any, was taken. Reuters did not identify the hacker.
Why It Matters
The economics are the story. A campaign that once required a skilled team working for weeks was, by this account, run by one person who wrote a few lines of instructions and paid for computing time. Fagerland told Reuters he had assumed criminal agents would still need human supervision and that the technology had moved past that. “It’s a new world now,” he said.
For boards and executives, the lesson is less about any single model than about exposure. The sites in this campaign were the kind organizations tend to forget: product support pages, marketing sites and pages operated by vendors. Leaders should know who owns every public-facing site, including those run by third parties, and how quickly a compromise would be detected.
The tools named are widely available commercial or open-source products, and nothing in the reporting indicates that their makers knew of the activity. Reuters’ article carried no comment from Anthropic or DeepSeek. Cairn’s author, who goes by “le1xia0” online, acknowledged a Reuters message without answering questions and has said on X that the platform was not designed for hacking.
Context
The case follows an earlier campaign against online retailers documented by Israeli firm Gambit Security. Reuters also cited security firm Bitsight as saying talk of autonomous agents is surging on criminal forums.
What to Watch
- Whether affected companies disclose that customer or business data was taken, and whether more victims come forward.
- Whether NorForge publishes its full findings, and whether other researchers confirm the costs and techniques described.
- How AI model providers detect and respond to criminal use of coding agents.
- Whether regulators or law enforcement agencies comment on the campaign.
