AEGLIX News
Amazon Threat Intelligence has disclosed that suspected Chinese-linked attackers leveraged an autonomous AI agent platform named Cairn to identify and exploit a zero-day vulnerability in a medical imaging application, ultimately gaining full control of a server. The operation was carried out by a Chinese-nexus threat actor from September 8 to 10, 2026, with a primary focus on East Asian educational institutions located outside of China. Amazon researchers detected the campaign after finding that the operators had left their Cairn instance exposed to the internet without authentication, which laid bare their hacking infrastructure.
Cairn functions as a legitimate penetration testing tool designed to orchestrate multiple AI agents for cyber operations, originally built to advance research and bolster defenses. However, the platform has been weaponized in real-world attacks. Gambit Security recorded instances of Cairn being abused in a large-scale credit card theft operation during September. Additionally, Google observed the framework deployed in an intrusion targeting a U.S. aerospace organization. In these scenarios, operators directed the AI agents to gain shell access, after which the agents autonomously discovered and exploited vulnerabilities.
Amazon reports that Cairn's AI agents uncovered a path traversal zero-day in MONAI Label, an open-source medical imaging platform that employs AI to assist researchers with labeling medical images. This flaw allowed the agents to write files to arbitrary server locations, and they progressed from API reconnaissance to full exploitation in just 11 minutes. The agents leveraged the vulnerability to plant a Python reverse shell in the sitecustomize.py file, which executes automatically upon Python startup, then invoked an API endpoint to secure root access. The campaign also revealed limitations: a 16-hour, 148-action attempt to compromise a Redis server failed, and in a separate university GPU server intrusion, the agents tested over 30,000 credential combinations across six hours without succeeding in a brute-force breach.
Amazon Threat Intelligence exposed the operation after detecting that the threat actors had left their Cairn deployment publicly accessible without any authentication, inadvertently revealing their attack infrastructure. The offensive activity was attributed to a Chinese-nexus actor and took place between September 8 and 10, 2026, directed mainly at East Asian educational institutions outside China.
Report prepared 2026-11-20 · Based on reporting from BleepingComputer. Source publication date above; subsequent developments may not be reflected.
Sources
Read original source · BleepingComputer (opens in a new tab)