MONAI Label Path Traversal Vulnerability Exploited by AI Agents
A China-linked threat actor used autonomous Cairn AI agents to discover and exploit a critical path traversal zero-day vulnerability in MONAI Label to achieve root-level code execution.
Triage facts
Revision 1- Applicability
- MONAI Label
- Evidence
- Confirmed activeHigh confidence
- Severity
- Critical
- Sources
- 1 public source
- Freshness
- Revision 1Nov 21, 13:04 UTC
- Source quality
- C2Likely
Overview
Amazon threat intelligence researchers identified a campaign in September 2026 where a China-linked threat actor utilized autonomous AI agents operating via the Cairn open source penetration testing framework to exploit a critical path traversal vulnerability in MONAI Label. The researchers discovered the activity after observing an unauthenticated instance of the Cairn framework.
Attack Chain
- Reconnaissance and Access: The threat actor utilized AI agents to target university networks in East Asia. The agents were instructed to obtain shell access.
- Vulnerability Discovery: The agents mapped the application API of MONAI Label and identified a path traversal bug in a file upload function.
- Exploitation: The agents used the vulnerability to write arbitrary files to the server. In one instance, the agents planted a Python reverse shell in a file that executes upon Python interpreter startup to achieve a root shell.
- Lateral Movement: The actor utilized a proxy on a compromised host to tunnel into a second network and target additional devices running the same vulnerable software.
Impact
Successful exploitation allows for arbitrary file writes and root-level code execution. The vulnerability enables any user who can reach the server over the network to exploit it, as authentication is disabled by default.
Recommendation
MONAI Label versions 0.8.5 and earlier are affected. A fix has been created and will be included in the next release of the software.