Hackers employed an autonomous AI agent platform named Cairn to uncover a zero-day flaw in a medical imaging application and used it to seize full control of a server, according to Amazon Threat Intelligence. The attackers gave the agents a single instruction, to obtain shell access, and the agents located and exploited the vulnerability on their own. Amazon's researchers found the operation only because the attackers left their Cairn instance reachable from the internet without authentication, exposing their activity. The campaign targeted mainly East Asian educational institutions outside China, according to the researchers.
Amazon attributed the attacks to a Chinese-nexus threat actor with moderate confidence, citing the wording of the retrieved prompts, the timing of the activity and the selection of targets. The activity ran between September 8 and 10, 2026, and Amazon published two IP addresses linked to the campaign that were observed during that window. Cairn itself is a legitimate penetration-testing tool built to coordinate multiple AI agents during cyber-operations, intended for research and defensive hardening. It has also surfaced in other abuse cases: Gambit Security documented its use in a large-scale credit card theft operation in September, and Google recently reported its deployment against a U.S. aerospace organization.
The agents did not succeed every time. In one attempt to take over a Redis server, a 16-hour effort spanning 148 actions failed. In another case aimed at a university GPU server, the agents spent six hours trying more than 30,000 credential combinations and never brute-forced their way in.
The successful intrusion came through a path traversal zero-day in MONAI Label, an open-source medical imaging application that uses AI to help researchers label medical images. The flaw let the agents write files to arbitrary locations on the server, and they moved from mapping the application's API to exploiting the vulnerability within 11 minutes. In a later session, according to Amazon, the agents used the flaw to write a Python reverse shell into the sitecustomize.py file that loads automatically when Python starts, then triggered an API endpoint to obtain root access. After the zero-day was found, the operator gave the agents a proxy into a second university network running the same vulnerable software and directed them at other machines.
Amazon reported the flaw to the MONAI Label maintainers, who issued a security update on November 18. The company warned that AI agents can locate and exploit vulnerabilities faster than security teams can respond, and urged organizations to lean on automated defenses rather than manual investigation.
More cybersecurity news from TechManNews.







