egov.mn
Монголоор унших

China-linked hackers' AI agents found a zero-day on their own and took over a server, Amazon says

Amazon Threat Intelligence says a suspected China-linked actor used the autonomous AI framework Cairn against East Asian universities, and its agents discovered and exploited a previously unknown flaw in medical imaging software to gain root access in minutes.

This story was prepared automatically by software.

Person using a laptop displaying code and an AI graphic in front of a Chinese flag
Person using a laptop displaying code and an AI graphic in front of a Chinese flag

A suspected China-linked threat actor used autonomous AI agents to carry out hacking operations with minimal human involvement, and the agents independently discovered and exploited a zero-day vulnerability to take full control of a server, Amazon Threat Intelligence said on November 20. "These findings show that agentic offensive operations are already happening," wrote CJ Moses, chief information security officer of Amazon Integrated Security.

An exposed server revealed the campaign

On September 10, Amazon found a server publicly hosting Cairn, an open-source autonomous AI penetration-testing framework, running without authentication. That exposed the operator's project configurations, prompts and AI output. The prompts were written in Simplified Chinese and script timestamps were set to China Standard Time. The campaign targeted IP addresses tied to universities in East Asia outside mainland China over roughly 48 hours beginning September 8.

The actor gave the agents a simple goal, variations of "obtain shell access," and the agents planned, executed and adjusted their approach without the operator choosing each step.

From API mapping to root in minutes

Against a server running an open-source medical imaging AI application (MONAI Label), the agents mapped the API, found a path traversal flaw in an upload endpoint and, in about 11 minutes, were able to write arbitrary files to the server. In a later session they used the flaw to plant a Python reverse shell that ran with root privileges. Amazon assessed with high confidence that the flaw was a zero-day; after it notified the maintainers, they released a fix and a security advisory.

Other operations failed but showed persistence: agents spent about 16 hours and 148 sequential actions trying to take over an open Redis server, and tested more than 30,000 credential combinations against a university GPU server and other systems.

Lessons for defenders

Amazon attributed the activity to a Chinese-nexus actor with moderate confidence. It said agents that find and exploit flaws in minutes and work around the clock compress the window for defense, so organizations should prioritize high-confidence detections that trigger automated responses, with humans overseeing outcomes rather than every action. Amazon also noted that Gambit Security has documented a separate, financially motivated group using Cairn against online retailers.

Sources: AWS Security Blog, The Register, GitHub Security Advisory

Share
Next storyTwo million planted secrets reveal AI supply chain leaking user prompts, Dragos researcher finds
Newsletter

Get the latest news delivered to your inbox.

Related articles