Back

Suspected Chinese spies used AI agents to find and exploit zero-day in university server
SiTech AI Team2 min read

Suspected Chinese spies used AI agents to find and exploit zero-day in university server

Amazon says suspected Chinese spies used autonomous AI agents to find a zero-day in MONAI Label and take control of a university server after earlier failed attacks.

Suspected Chinese spies used autonomous AI agents to find and exploit a zero-day vulnerability in the open source MONAI Label medical imaging application and gain full control of a university server, according to Amazon CISO CJ Moses. Amazon's threat intelligence team found a publicly hosted instance of the Cairn penetration-testing framework on September 10, giving visibility into a campaign targeting universities in East Asia outside Mainland China. Moses said the agents kept adapting for nearly a full day without human direction.

Failed intrusions before the zero-day

Before finding the flaw, the agents spent about 16 hours and 148 sequential actions trying to break into another university's Redis server, including attempts to load unauthorized code, escape the scripting sandbox, write configuration files and corrupt memory. Those attempts failed, as did a Log4Shell injection, a server-side request forgery attack against a Qdrant instance, and credential sprays against PostgreSQL, SSH and other tools. A separate operation lasting more than six hours targeted a university GPU server with a credential spray and compiled 80 real usernames from public websites, academic papers and other documents.

How the zero-day was exploited

The agents then found a path traversal vulnerability in MONAI Label, mapped its API endpoints, and tested an upload endpoint's filename parameter. That parameter was vulnerable, and in about 11 minutes the agents went from mapping the API to arbitrarily writing files on the server. Later they used the same flaw to execute a Python reverse shell running as root, and in another session a human operator gave them a proxy that tunneled into a second university network connected to a server running the same vulnerable code. Amazon notified the project's maintainers, who released a fix and a security advisory.

Broader pattern of autonomous AI attacks

The campaign ran for 48 hours beginning September 8, with prompts written in Simplified Chinese and timestamps in executed scripts configured to China Standard Time. Amazon's discovery followed an earlier Gambit Security report that documented a Chinese operator using Cairn in at least 105 attacks, compromising at least 27 companies to varying degrees. Moses said the cases show the autonomous AI agent framework is being adopted in both espionage and criminal operations.

Sources: The Register

SiTech

SiTech — AI-powered web development

We build fast, modern websites and bring AI into real business workflows. Have a project or a question? We'd love to help.