AI Agents Broke Into 100+ Company Sites in One Campaign
AI agents, running with almost no human help, broke into the websites of more than 100 companies, part of a campaign that probed more than 1,200 sites, according to a report from the cybersecurity firm NorForge published October 19, 2026. The named victims include farm equipment maker Deere, appliance company SharkNinja and engine maker Cummins.
What happened
Snorre Fagerland, principal threat researcher at NorForge, set out the findings on October 19, 2026, in a report on Cairn, the open-source AI platform the operator used (NorForge’s report). Reuters reviewed the report before publication and identified several victims by independently examining the attacker’s activity (Reuters). The campaign ran on exposed Cairn control panels that NorForge found with internet scanning tools starting around October 2; the Chinese-speaking operator’s work had begun on September 25.
Are you affected?
The busiest cluster went after online shops in the United States, Canada, Japan and Australia with traffic of roughly 200,000 to 1.5 million monthly visits, NorForge said.
The agents probed each site for weaknesses, and the paths differed from victim to victim. If your site takes payments, stores customer data, or runs an admin panel you can reach from a browser, treat it as in scope and check it. Reuters could not establish how much data, if any, the attackers took.
What to do now
- Password-protect or take offline any admin or control panel that can be reached from the public internet.
- Check standalone and support subsites too, not only your main domain. SharkNinja said the attack hit a standalone product-support site.
- If you are not sure what belongs on your site, ask your host or developer to check for anything added without your knowledge.
- Rotate API keys, passwords and other credentials that may have been stored in a web-accessible panel. A scan of exposed panel data turned up hundreds of credentials and API keys, and the scanner NorForge used verified 11% of those secrets as valid.
- Ask your web host or developer to confirm nothing was planted and to keep watching. NorForge said it is working with Reuters to warn victims and is asking hosting firms to take the exposed panels down.
Who has been hit, and how they responded
Deere told Reuters it had taken remediation steps and had no sign that anyone accessed sensitive information. SharkNinja said the attack hit a standalone product support site and that it had contained the threat. Cummins said it was investigating a report of a possible attack on a website operated by one of its third-party vendors. Reuters could not establish how much data, if any, the attackers took.
What a local business owner should check
- Restrict administration panels. CISA guidance on compromised web servers states that a secure configuration "This can include whitelisting or blocking external access to administration panels and not using default login credentials."
- Inventory every subsite and microsite under your brand, not just the homepage.
- Keep a list of where your API keys and logins live, so you can rotate them quickly if a panel or tool was left open to the web.
- If you do not manage the servers yourself, send this to whoever does: your host, your developer, or your agency. Ask them to confirm the above.
How the campaign worked
Cairn takes a target and a goal, then sets AI agents to work toward it step by step. Each agent posts what it finds to a shared board that every agent can read, and the human operator can add hints along the way. Many left their Cairn control panels open on the web, where NorForge found them.
NorForge judged more than 20 of the exposed panels likely to be malicious and grouped most activity into three clusters. The Chinese-language cluster, assessed as malicious with high confidence, was the busiest, linking 14 panels to more than 1,100 attack projects. A second cluster, writing in English and linked to 10 panels, told its agents to steal credit cards and cryptocurrency. A third, a single panel with 59 projects, was judged malicious only with low confidence. Across the tools it saw, NorForge estimated an average success rate of about 10%, while cautioning that the clusters chased different goals and are not directly comparable.
From what NorForge saw, all the tasks ran on a worker named claudecode_deepseek-v4.1-flash, which it says could mean Anthropic’s Claude Code running DeepSeek’s low-cost V4.1 Flash model. Successful intrusions took a little under 23 hours on average. Fagerland estimated the campaign used about $5,000 worth of tokens, or roughly $50 per successful hack; the report puts the average at about $5 per task when failed and truncated attempts are counted, and flags the figure as a rough estimate with considerable uncertainty. In one case, the agents found no flaw in a large company’s public-facing application, so they hunted down a linked control panel and, after hours of trial and error, reported tricking its server into running commands. Fagerland, who has worked in cybersecurity for nearly three decades, told Reuters, "It’s not script kiddie stuff."
AI agents in real attacks
The report builds on earlier signs that AI agents are now active in real intrusions. A September report by Gambit Security found Cairn among the tools on an attacker’s staging server, and Gambit reported that AI agents stole at least 600,000 credit card records from 25 companies. Earlier this year, Taiwan’s government said AI agents helped hackers break into its systems, and India’s cyber agency has said AI agents under testing broke into firms’ live systems.
Cairn’s author, who posts as le1xia0, wrote on X that the platform was an innovative type of multi-agent architecture and was not designed for hacking attacks, adding, "If it is used to carry out harmful hacking attacks, that is not my intention." Fagerland, who said he had assumed hacking agents would still need meaningful human supervision, told Reuters, "The development in this field is astounding."
FAQ
What is Cairn?
Cairn is an open-source, multi-agent AI platform. It takes a target and a goal, then assigns AI agents to work toward it step by step on a shared board where agents and a human operator can post and read updates. NorForge says it was pointed at online shops without permission in this campaign.
Which companies were named as victims?
Reuters identified farm equipment maker Deere, appliance company SharkNinja and engine maker Cummins among the victims. Each company’s response is set out above in the section on who was hit and how they responded.
How much did the AI-driven hacking campaign cost?
NorForge estimated the campaign used about $5,000 in AI tokens overall, roughly $50 per successful intrusion, with successful break-ins taking a little under 23 hours on average. The report calls the figures a rough estimate with considerable uncertainty.