Cairn AI Agents Found a MONAI Label Flaw and Took Over a University Server

Brendan Smith
Brendan Smith - Cybersecurity Analyst
3 Min Read
A scanner shines three red beams through a gap in a server beside the words AI agents found a flaw.
Amazon traced the university server compromise to AI agents that found a flaw in MONAI Label.

Amazon Threat Intelligence reported on November 20 that AI agents found a previously unknown flaw in MONAI Label and helped take control of a university server in East Asia. MONAI Label helps research teams annotate medical images. Version 0.8.6, published to PyPI on November 22 UTC, fixes the flaw. Teams running 0.8.5 or older should update and investigate any server that was reachable from an untrusted network.

How Amazon reconstructed the September attack

On September 10, Amazon found the attacker’s Cairn control panel on the open internet without a login. Its records showed what the operator asked the AI agents to do and how the agents responded. The campaign had begun roughly two days earlier and targeted university networks outside mainland China.

The records also showed where automation failed. The agents spent roughly 16 hours trying to gain control of one database server without succeeding. Attempts to log in to a research computing server also failed. On the MONAI Label server, the outcome changed when a flaw let the agents write files outside the application’s normal storage area. Amazon measured about 11 minutes from the initial examination of the application to that file write. A later session gained administrator-level control of the server.

Amazon inferred that the operator abused the same flaw to get into a second university network. The report leaves the universities unnamed and makes no finding about whether patient or research data was taken. Our earlier report on exposed Cairn panels covers separate investigations into use of the same tool.

MONAI Label 0.8.6 is available

The November 18 advisory identified the affected versions. The November 22 release supplies a packaged fix for CVE-2026-133968. The maintainers say that releases through 0.8.5 are affected and that the server ships with authentication off by default. Check the version actually running on the server, including older containers or lab instances.

  • If an older server is still reachable from an untrusted network, restrict access now. Enable authentication, limit user roles and run the service with access only to the storage it needs.
  • If the server was previously exposed, review it before returning it to service. Preserve logs and have the incident-response team check for unexpected file changes or signs of unauthorized access. The update closes the flaw. Investigating earlier access still requires evidence from the server.

References

  1. Moses, C. Amazon’s investigation into the Cairn attacks. AWS Security Blog, November 20, 2026.
  2. Project MONAI. MONAI Label 0.8.6 release notes. GitHub, November 22, 2026 UTC.
  3. Python Package Index. MONAI Label 0.8.6 package. November 22, 2026 UTC.
TAGGED:
Share This Article
Cybersecurity Analyst
Follow:
Brendan Smith has spent over 15 years knee-deep in cybersecurity, chasing down malware from the gritty reverse-engineering of old-school trojans all the way to wrangling full-blown incident responses for small-to-medium businesses that couldn’t afford a full-blown breach. Over at Gridinsoft, he’s the guy piecing together those double-checked guides on nasty stuff like AsyncRAT, a remote access tool used in malware campaigns—helping readers make sense of the threat and work through cleanup without the extra headache.
Leave a Comment

AI Assistant

Hello! 👋 How can I help you today?