Amazon: China-Linked Hackers’ AI Agents Exploited Medical Software Zero-Day

Amazon has reported that a China-linked threat actor used autonomous AI agents to discover and exploit a zero-day vulnerability in MONAI Label, an open-source tool for AI-assisted annotation of medical images. Source: securityweek.com.

What Happened

Amazon’s Threat Intelligence team first detected the attack in September. It began with an exposed instance of Cairn, an open-source AI penetration testing framework. The instance had no authentication, so anyone could see the operator’s prompts and the agents’ outputs.

The operator listed universities in East Asia—excluding mainland China—as targets. The AI agents were told to scan these systems and gain shell access. After scanning one, they found a path traversal flaw in MONAI Label’s file upload function.

Within 11 minutes, the agents moved from mapping the API to writing arbitrary files on the server. This shows how quickly AI agents can act in cyber operations. In a later session, they planted a Python reverse shell in a file that runs at startup. This gave them full root access to the system.

Key Facts

How the Vulnerability Works

MONAI Label helps medical professionals label X-rays, MRIs, and CT scans. It allows users to upload images for AI processing. This is a core part of its workflow.

The path traversal flaw occurs when the software fails to validate file paths during uploads. An attacker can manipulate the path to access or write files outside the intended directory. Because authentication is disabled by default, no login is needed to exploit it.

Once triggered, attackers can upload and execute malicious code. This could lead to full system compromise—such as installing backdoors or stealing sensitive data.

Why It Matters

Medical imaging tools handle highly sensitive patient data. A breach could expose personal health records, leading to privacy violations and fraud risks.

Artificial Intelligence (AI). Via original publisher. · Source

Open-source tools are popular for their flexibility and cost. But this incident shows a serious risk: default insecure settings can be exploited in large-scale, automated attacks. The use of AI agents to autonomously find and exploit the flaw marks a growing trend in offensive cyber operations.

Limitations and Open Questions

Amazon has not confirmed the threat actor’s identity. It only attributes the activity to a ‘Chinese-nexus’ group with moderate confidence. This is based on prompt language, operation time zones, and the strategic value of the targeted institutions.

No evidence shows patient data was stolen or real-world harm occurred. Still, the incident raises serious concerns about open-source tool security and AI-driven attack methods.

The MONAI Label developers released a critical advisory. The fix addresses the path traversal flaw and improves authentication and input validation. It will be included in the next software release.

What to Watch Next

Users of MONAI Label should check their version. They must ensure authentication is enabled and properly configured. Regular audits of open-source tools are essential—especially those with default insecure settings.

Security teams should monitor for AI-driven attacks using autonomous agents in penetration testing tools. This event aligns with recent AI-assisted cyberattacks reported by the European Commission and others.

For more on similar threats, see China-Linked Hackers Exploit Critical VMware vCenter Zero-Day and Hackers Exploit PHP2Shell Flaws in Check Point Gateways.

Sources & further reading

Featured image: Artificial Intelligence (AI). Via original publisher. Image source

Exit mobile version