Canadian Cyber Security Journal
SOCIAL:
Filed under: Trends

AI Agents Found and Exploited a Medical Imaging Zero-Day in 11 Minutes: What Canadian Health and Research Teams Must Do

What Happened

A Cairn AI agent zero-day attack shows how fast automated tools now move from reconnaissance to root. Amazon Threat Intelligence reports a suspected China-nexus operator used Cairn, an open-source AI agent platform built for penetration testing, against MONAI Label, an open-source medical imaging application, BleepingComputer reports.

The operator gave the agents one goal: obtain shell access. The agents mapped the API, found a path traversal flaw, and exploited it in about 11 minutes. They wrote a Python reverse shell into sitecustomize.py and triggered an API endpoint to get root. The operator then pointed the agents at a second university network running the same software. The activity ran from September 8 to 10 and targeted East Asian educational institutions. Amazon spotted the campaign because the attackers left their Cairn instance open without authentication. MONAI Label maintainers shipped a fix on November 18 under advisory GHSA-qc22-rmgc-vg94.

Why This Matters for Canadian Organizations

Canadian teaching hospitals, cancer centres, and university research groups run open-source medical imaging and AI tools on GPU servers. Many of these servers sit on research networks with lighter controls than clinical systems. They often hold patient imaging data under PHIPA, provincial health privacy laws, and PIPEDA. They also fall within federal research security expectations.

The bigger lesson is speed. Your patch cycle measures in days or weeks. These agents found and used an unknown flaw in minutes. The attempts did not all succeed. A 16-hour attack on a Redis server failed. Still, the operator only needed one win. Pair this with today’s Bell Canada agentic breach and the trend is clear. Automated attackers now target Canadian infrastructure.

What to Do

Update MONAI Label to the fixed release now. Inventory every open-source research application exposed beyond a lab network, and put those APIs behind a VPN or identity-aware proxy. Watch for unexpected writes to Python startup files such as sitecustomize.py. Block the indicators Amazon published, 142.4.98.145 and 142.4.107.34. Rate-limit and alert on rapid API enumeration. Invest in automated detection and response, since human triage alone will not match agent speed. Follow our Trends coverage and today’s Daily Brief for updates.

Enjoy this article? Don’t forget to share.