Attackers used Cairn AI agents to discover an unknown MONAI Label flaw and take full control of a server, Amazon Threat Intelligence reported. The campaign ran from September 8 to 10, 2026, mainly targeting educational institutions in East Asia outside China.
MONAI Label is an open-source application that helps researchers label medical images using AI. Its maintainers released a security update on November 18 after Amazon disclosed the vulnerability.
Amazon said the operators told the agents to “obtain shell access” without specifying how. The agents then found and exploited the flaw independently.
The vulnerability involved path traversal, which allowed files to be written to arbitrary locations on the server. The agents took 11 minutes to move from mapping the application's API to exploiting the bug.
In a later session, they placed a Python reverse shell in sitecustomize.py, a file loaded automatically when Python starts. They then triggered an API endpoint and gained root access, giving them full control of the server.
The operator subsequently supplied a proxy into a second university network using the same vulnerable software. The agents were instructed to attack other machines there.
Amazon uncovered the activity because the attackers had exposed their Cairn instance online without authentication. That mistake gave researchers access to records of the operation, according to Amazon's investigation.
Cairn is a legitimate penetration-testing platform designed to coordinate multiple AI agents for research and defensive security work. Its use did not guarantee success in the other attacks Amazon examined.
An attempt against a Redis server failed after 16 hours and 148 actions. At a university GPU server, the agents tried over 30,000 credential combinations across six hours without gaining access.
Amazon assessed a Chinese connection with moderate confidence. Researchers based that judgment on prompt language, activity timestamps and the targets selected.
The company also published two IP addresses associated with the campaign during September 8–10, 2026:
142[.]4.98.145142[.]4.107.34
Other researchers have also reported misuse of Cairn. Gambit Security documented its role in a large-scale credit card theft operation in September. More recently, Google reported its deployment against a U.S. aerospace organization.
Amazon warned that AI agents can discover and exploit flaws faster than security teams can respond. It urged organizations to place greater reliance on automated defenses rather than manual investigations.

