AI agents went from an API map to a root shell in 11 minutes. No patch is out yet.
By Lior AharonovFounder, 7IT Solutions ·
What happened
Amazon says attackers used AI agents from Cairn, an open source penetration testing framework, against universities in East Asia. They abused a path traversal flaw in MONAI Label, a medical image annotation tool, and reached a root shell. The advisory covers 0.8.5 and earlier and lists no fixed release.
Why it matters
Authentication is off by default, so anyone who can reach the server can write files on it. Picture a research box left open for a weekend demo, and an attacker who is not a person typing but agents that needed about 11 minutes.
What to do
If you run MONAI Label 0.8.5 or earlier, turn authentication on and put it behind a network allowlist today, then watch the project for the next release, which carries the fix.

My take
Lior Aharonov
Founder, 7IT Solutions
The agent did not need a clever bug. It needed a default with no login and about eleven minutes.
Sources
| SecurityWeek: China-linked hackers used AI agents to exploit a medical software zero-day | securityweek.com |
| MONAI Label advisory GHSA-qc22-rmgc-vg94 | github.com |
Researched with AI tools; every fact is checked against the linked sources.