News
Open-Source AI Agent Platform Cairn Abused in Attacks, Reportedly Hitting 100+ Companies
Security firm Norforge describes at least three apparently malicious clusters of activity that use the open-source Cairn platform to automate attacks on online stores and payment services. According to Reuters, one attacker breached more than 100 companies, including John Deere and SharkNinja.
Contents
Cybercriminals are using Cairn, an open-source platform that coordinates AI agents, to automate break-ins on websites. AI security firm Norforge identified at least three clusters of such activity, and according to Reuters the largest one led to breaches at more than 100 companies.
What Cairn is
Cairn is a general-purpose problem-solving engine that steers AI agents toward a set goal. It can work as an autonomous penetration testing engine. Each task is a "project" with an origin and a predefined goal, and the agents move step by step from one to the other, posting their intents and findings to a shared blackboard.
Norforge came across the case while examining an earlier Gambit Security report about hackers using three AI tools to automate attacks on hundreds of online stores. Some of the attackers left their Cairn dashboards exposed to the internet, which let researchers look at data on the projects they were running.
Three clusters of attackers
The largest cluster operates in Chinese. It has more than 1,100 attack projects across 14 dashboards and targets e-commerce and payment infrastructure in the United States, Japan and Australia. According to Norforge, its goal is persistent access to large platforms through webshells, probably with the intent to monetize it later. Targets were ranked by traffic and automatically validated with Microsoft's Playwright framework before being fed to Cairn.
The second cluster uses English and has 10 exposed dashboards. It focuses on stealing payment cards and cryptocurrency, mainly from merchant payment services, gray-market payment sites and pharmaceutical platforms. Some of its targets are African commerce sites, and Norforge believes the actor is based in Africa. That is the firm's assessment, not a confirmed finding.
The third cluster is a single exposed dashboard with 59 projects aimed at various online stores, including very large US platforms. It is interested in shell access, as well as admin panels and databases.
Success rate and victims
According to Norforge, the attackers succeeded in roughly 10 percent of their attempts. Hundreds of secrets were found in the dashboard data, and 11 percent of them were verified as valid. Norforge does not name the affected companies. Reuters reports that the actor behind the largest cluster breached more than 100 companies, including US equipment maker John Deere and home appliance company SharkNinja.
SecurityWeek notes that the victims' names come from Reuters, not from the Norforge report. The text also gives no dates for the attacks themselves. The activity of the described clusters appears to be malicious, and placing the second cluster in Africa is the researchers' assessment.
What it means for businesses
The case shows that publicly available agent-coordination tools lower the barrier to automated scanning and breaking into websites at the scale of hundreds of targets. For Polish online stores and payment operators, the takeaway is practical: poorly secured sites and admin panels can be probed en masse and largely automatically. Still, it was the attackers' own Cairn dashboards, left exposed to the internet, that helped researchers track their activity.
