Attackers' AI Agents Found a Zero-Day in Medical-Imaging Software and Gained Root Access

What happened
Amazon Threat Intelligence reported that autonomous AI agents, believed to be controlled by China-linked hackers, exploited a previously unknown vulnerability in MONAI Label, an open-source medical imaging server. The attackers used the flaw to gain root access on a university server in September 2026. The vulnerability allowed unauthenticated users to write files outside the intended storage folder, and the AI agents mapped the server and executed the attack within minutes. Although a fix has been merged into the source code, there is currently no patched release available for users to install.
Key details
- The vulnerability affects MONAI Label versions 0.8.5 and earlier.
- Authentication is off by default in MONAI Label, allowing unauthenticated access.
- The flaw is now tracked as CVE-2026-133968 and rated 9.8 out of 10 in severity.
- A fix was merged into the source code on October 19, 2026, but no new release is available yet.
- The AI agents used the Cairn framework and reached file-write capability in about 11 minutes.
Why it matters
Research teams and universities using MONAI Label are at risk of unauthorized access and potential server compromise until a patched version is released. Administrators must take immediate steps to block untrusted network access, enable authentication, and restrict server permissions to reduce exposure. The absence of an installable fix means affected organizations face ongoing risk and must rely on workarounds to protect sensitive systems.
