Cairn: Open windows into AI-driven cyber operations

NorForge investigation uncovered more than 20 publicly exposed Cairn control panels and used those panels to reveal how Cairn-powered operations plan, automate, and execute AI-driven intrusions. Cairn is a state-based problem-solving engine that runs projects from an origin to a defined goal, recording discovered facts and planned intents on a shared project blackboard that both agents and human operators can read. The exposed dashboards leak highly sensitive operational data - target inventories, vulnerability findings, credentials, API keys and evidence of successful compromises - and even legitimate red-team engagements were found left unsecured, putting clients at risk. Panel addresses were redacted to prevent further abuse.

One prominent cluster, AIA0001, operates in Chinese across 14 panels with over 1,100 attack projects targeting e-commerce and payment infrastructure in the US, Japan and Australia for monetization, and is assessed as financially motivated rather than state-sponsored. Operators feed targets through an automated pipeline that validates sites with Playwright and ranks targets by traffic; an observable domain-normalization bug (dropping initial “w”) helped link panels. Worker directories exposed comprehensive tooling - Playwright/Chromium, PHPGGC, StyleSmuggler PoC, Magento validation labs, Docker, Python, Node.js, Nmap, ffuf, Amass and more - and artifacts show reuse of a Tencent CTF setup. Agents ran as claudecode_deepseek-v4.1-flash; estimated task costs average around $5, with long-running projects exceeding $100.