{"id":2006,"date":"2026-10-19T19:09:06","date_gmt":"2026-10-19T19:09:06","guid":{"rendered":"https:\/\/seoscanpro.ai\/blog\/ai-agents-hacking-campaign-cairn-100-sites\/"},"modified":"2026-10-24T09:39:41","modified_gmt":"2026-10-24T09:39:41","slug":"ai-agents-hacking-campaign-cairn-100-sites","status":"publish","type":"post","link":"https:\/\/seoscanpro.ai\/blog\/ai-agents-hacking-campaign-cairn-100-sites\/","title":{"rendered":"AI Agents Break Into 100+ Company Sites in Mostly Autonomous Hacking Campaign"},"content":{"rendered":"
Updated October 22, 2026 at 3:40 AM ET<\/em><\/div>\n

According to Reuters, agents had tried to break into more than 1,200 sites by late last week. NorForge also said a scan found hundreds of credentials, API keys and other secrets in exposed panels.<\/p>\n

Reuters identified some victims, including Deere, SharkNinja and Cummins, while TNW said it had not independently verified the intrusions.<\/p>\n

What happened<\/h2>\n

Cairn operators gave AI agents targets and goals. The agents worked through the tasks, shared findings on a common board and could receive hints from a human operator. NorForge assessed one Chinese-language cluster as malicious with high confidence and found more than 1,100 projects across 14 linked panels.<\/p>\n

The agents tried to run code on online shops and plant an uploader web shell. NorForge checked one claimed break-in at a large hardware parts vendor and found the shell live. By late last week, the agents had tried to enter more than 1,200 sites, according to Reuters.<\/p>\n

Are you affected?<\/h2>\n

The activity targeted online shops with reported traffic of 200,000 to 1.5 million in the United States, Canada, Japan and Australia. NorForge said the apparent goal was lasting access to shop and payment systems.<\/p>\n

The exposed panels also contained secrets collected from targets. A scan found hundreds of credentials, API keys and other secrets. Trufflehog verified 11% of the exposed secrets as valid. Some security testing firms had also left client work exposed.<\/p>\n

NorForge has not listed all affected websites or confirmed that every attempted intrusion succeeded. Deere said 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 possible attack on a website run by a vendor. Reuters did not establish how much data, if any, was taken.<\/p>\n

What to do now<\/h2>\n
    \n
  1. \n

    Check your web server for unauthorized uploaded files, web shells, unexpected scripts and signs that the application is running commands it should not run.<\/p>\n<\/li>\n

  2. \n

    Review shop, payment and support-system logs for unusual account activity, new administrative access and unauthorized file changes.<\/p>\n<\/li>\n

  3. \n

    Rotate credentials and API keys that appeared in exposed panel data, then update any systems that used them.<\/p>\n<\/li>\n

  4. \n

    Ask your hosting provider about any security concerns associated with your online services.<\/p>\n<\/li>\n

  5. \n

    Ask hosting and security partners whether they found a web shell, credential exposure or unauthorized access on your systems.<\/p>\n<\/li>\n<\/ol>\n

    What does this mean for a website audit?<\/h2>\n

    NorForge estimated that successful attacks took a little under 23 hours on average. The campaign used about $5,000 in tokens, or roughly $50 per successful hack, according to Snorre Fagerland, principal threat researcher at NorForge. The figure is an estimate.<\/p>\n

    NorForge put the average at about $5 per task when failed and early-ending attempts were included. Across the tools it observed, NorForge estimated an average success rate of about 10%. The researchers said the clusters had different goals and could not be compared directly.<\/p>\n

    NorForge<\/a> also states that projects with more than 1,000 intents and facts could cost more than $100.<\/p>\n

    How did the agents work?<\/h2>\n

    Cairn is an open-source, multi-agent AI platform. It takes a target and a goal, then coordinates agents step by step. Agents post findings to a shared board, where other agents can read them and the human operator can add hints.<\/p>\n

    All observed tasks in the campaign ran through a worker named claudecode_deepseek-v4.1-flash. NorForge said the name could indicate Claude Code running DeepSeek’s V4.1 Flash model. The campaign’s agents were told to find a way to run code on each site and then plant an uploader web shell.<\/p>\n

    In one example, the agents found no flaw in a large company’s public-facing application. They searched for a linked control panel and reported that they had tricked its server into running commands. NorForge’s broader findings covered three clusters. One targeted online shops, one focused on stealing credit cards and cryptocurrency, and a third was assessed as malicious with low confidence.<\/p>\n

    Why do exposed control panels matter to site owners?<\/h2>\n

    The panels contained more than attack instructions. They also exposed collected secrets from target systems. According to NorForge<\/a>, a scan of the panel data found hundreds of credentials, API keys and other secrets, and Trufflehog, the scanner it used, verified 11% of the exposed secrets as valid.<\/p>\n

    Operators and security teams should review their own exposure of secrets, rotate any that may have leaked, and follow any guidance their hosting or security partners provide.<\/p>\n

    What did other security reports say?<\/h2>\n

    NorForge’s search followed a September report from Gambit Security. Gambit found Cairn among the tools on an attacker’s server and reported that AI agents stole at least 600,000 credit card records from 25 companies.<\/p>\n

    Earlier in 2026, Taiwan’s government said AI agents helped hackers break into its systems. India’s cyber agency reported that AI agents under testing broke into firms’ live production systems.<\/p>\n

    FAQ<\/h2>\n

    What is Cairn, and why is it central to this story?<\/h3>\n

    Cairn is an open-source multi-agent AI platform that takes a target and a goal, then coordinates AI agents that share findings on a common board while a human operator can add hints. NorForge published its findings after finding operators running attack campaigns through it.<\/p>\n

    Which companies were named as victims?<\/h3>\n

    Reuters identified some victims by reviewing the operator’s activity, including Deere, SharkNinja and Cummins. The wider campaign involved claims of more than 100 company sites and attempts against more than 1,200 sites.<\/p>\n

    How much did each successful hack cost?<\/h3>\n

    NorForge estimated that the campaign spent roughly $5,000 in tokens in total, with about $50 per successful intrusion once successes were isolated. Including failed and abandoned attempts, the average was about $5 per task. NorForge also said longer projects with more than 1,000 intents and facts could cost more than $100.<\/p>\n

    This article reports on publicly available information from the sources listed below. We have no reason to believe it is inaccurate, but we have not independently verified every claim, and details may change as the story develops.<\/em><\/div>\n
    \n

    Sources<\/h2>\n