Hacker used AI agents to target 100+ sites, researcher says
A Chinese-speaking operator leveraged the open-source Cairn platform to automate attacks on over 100 corporate websites, according to a new report from NorForge.
Key points
- NorForge identified a campaign using AI agents to target more than 100 websites.
- Victims include Deere, SharkNinja, and Cummins, according to Reuters reporting on the findings.
- The attacker used the Cairn platform, which automates tasks like finding code execution paths.
A Chinese-speaking hacker targeted more than 100 corporate websites by delegating nearly all technical tasks to artificial intelligence agents, according to a report published Monday by NorForge. Principal threat researcher Snorre Fagerland detailed the findings in a study of Cairn, an open-source AI problem-solving platform that the attacker exploited to automate the intrusion process. Reuters reviewed the report before its public release and identified several high-profile victims by analyzing the hacker’s digital activity logs.
The campaign targeted online retail platforms in the United States, Canada, Japan, and Australia. The attacker’s automated pipeline instructed AI agents to locate vulnerabilities that allowed code execution on these sites. Once access was gained, the agents deployed web shells, which are forms designed to save uploaded files rather than execute commands directly. NorForge assessed that the primary objective was establishing persistent access to shop and payment systems for future exploitation.
How the AI Platform Was Exploited
Cairn operates by accepting a specific target and goal, then deploying multiple AI agents to work toward that objective step by step. Each agent posts its discoveries to a shared digital board, allowing other agents and the human operator to read the findings and provide hints. This collaborative structure enables the system to solve complex problems autonomously. The human operator sets the initial targets and goals but can intervene with guidance as the agents progress through their assigned tasks.
NorForge researchers discovered many operators had left their Cairn control panels exposed on the public internet. Fagerland began identifying these open panels using internet scanning tools around October 2. The Chinese-speaking hacker’s specific campaign started on September 25. NorForge judged more than 20 of the discovered panels to be likely malicious. Most of the observed activity fell into three distinct clusters, with the Chinese-language cluster being the most active and assessed as malicious with high confidence.
Affected Organizations and Response
Reuters identified Deere, SharkNinja, and Cummins among the victims after reviewing the hacker’s activity records. Deere told Reuters it had taken remediation steps and found no evidence that sensitive information was accessed. SharkNinja stated the attack targeted a standalone product support site and that the company had contained the threat. Cummins said it was investigating a report of a possible attack on a website operated by one of its vendors. Reuters could not determine how much data, if any, the hackers successfully exfiltrated.
The Chinese-language cluster linked to 14 panels and contained more than 1,100 attack projects. By late last week, the group’s agents had attempted to break into more than 1,200 sites. NorForge verified one claimed breach at a large hardware parts vendor and confirmed the web shell was live. The typical targets were online shops with traffic volumes ranging from 200,000 to 1.5 million visits.
What should you do?
Organizations should monitor for unauthorized access to administrative interfaces and AI platform control panels. Security teams should review logs for signs of automated scanning or unusual file uploads, particularly web shells. If you use AI-assisted development or testing platforms, ensure they are not exposed to the public internet without strict authentication. No fix has been confirmed yet for the platform vulnerabilities.
What to do and how to stay safe: Cairn
- Audit your network for any exposed administrative panels or AI platform interfaces that may be accessible from the public internet.
- Review server logs for indicators of automated reconnaissance or the presence of unknown web shells and uploaded files.
- Implement strict access controls and multi-factor authentication for all systems that allow code execution or file uploads.
- Monitor for unusual outbound traffic that could indicate data exfiltration or communication with command-and-control servers.
Step-by-step guide: Social Engineering Warning Signs and How to Respond
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
Which companies were confirmed as victims in this AI-driven attack campaign?
Reuters identified Deere, SharkNinja, and Cummins as victims after reviewing the hacker's activity logs and reports.
What is the Cairn platform and how was it used in these attacks?
Cairn is an open-source AI problem-solving platform that uses agents to work toward a goal; the hacker used it to automate finding vulnerabilities and deploying web shells.
Did the attacker successfully steal sensitive data from the compromised sites?
Reuters could not establish how much data, if any, the hackers took; Deere reported no sign of sensitive information access.
