The Daily Byte
Breaking
Welcome to The Daily Byte — real stories from across Pakistan.
Saturday, 21 November 2026 · Pakistan

The Daily Byte

Stories from Pakistan that deserve to be heard

Cairn AI agents find zero-day, seize university server

By The Daily Byte ·21 Nov 2026 ·4 min read
Server racks in a university supercomputer hall at Linköping University

Hackers used autonomous AI agents to find a previously unknown vulnerability in a medical imaging application and exploit it to take complete control of a university server, according to Amazon Threat Intelligence. The agents went from mapping the application's interface to writing files anywhere on the server in about 11 minutes. The suspected Chinese-linked attackers had given the agents a single instruction, to obtain shell access, and left them to work out the rest on their own.

The campaign ran between 8 and 10 September 2026, targeting mainly educational institutions in East Asia outside mainland China. Amazon uncovered it after discovering that the attackers had left their Cairn instance, the open-source AI penetration-testing framework running the agents, publicly accessible on the internet without authentication. Because the server was open, its operational history was visible to anyone, including project configurations, prompts and the agents' output.

How the agents found the flaw

The target was MONAI Label, an open-source medical imaging application that uses AI to help researchers label medical images. The agents identified the application's framework, mapped its API endpoints and found an upload endpoint with a filename parameter. They tested that parameter for path traversal, a flaw that lets an attacker escape the intended directory, and it was vulnerable. Within 11 minutes of mapping the API, the agents were writing files to arbitrary locations on the server.

In a later session, the agents used the same flaw to write a Python reverse shell into sitecustomize.py, a file Python automatically loads when it starts, then triggered a training API endpoint to spawn a new Python process. The new process imported the payload and established a reverse shell running as root, giving the attackers full control of the server. Amazon says it assesses with high confidence that the flaw was a zero-day, because no advisory, fix, issue or prior mention of it existed for the project when it was found.

After the agents discovered the flaw, the human operator stepped back in. The operator gave the agents a proxy that tunnelled into a second university network, where a server was running the same vulnerable software, and directed them to target other machines there. Amazon reported the flaw to the MONAI Label maintainers, who released a security update and a GitHub security advisory on 18 November. The affected organisations were not Amazon customers, so Amazon worked with a government partner to notify them.

The attempts that failed

Not every operation in the campaign succeeded, and the failures are as revealing as the success. The longest operation targeted a Redis database server at one university that was configured to allow access without authentication. Over roughly 16 hours and 148 sequential actions, the agents tried a long list of known techniques, including abusing Redis replication to load unauthorised code, escaping its scripting sandbox, writing to configuration files and corrupting its memory, and even tried Log4Shell injection against another service on the same system. When those attempts failed, they mapped the internal Docker network through a server-side request forgery flaw in a co-located Qdrant instance and sprayed credentials against PostgreSQL, SSH and other services. None of it worked.

A separate operation ran for more than six hours against a university GPU server, a group of high-performance research computers. The agents built a list of about 80 real usernames from the university's public presence, including faculty directories, laboratory websites and academic papers, then tested more than 30,000 credential combinations against the server's remote login and other systems across the network. They never found a working login. Amazon's point is that the agents kept adapting for nearly a full day without human direction, while the one operation that found a new flaw needed only minutes.

Attribution and the wider warning

Amazon attributes the campaign to a Chinese-nexus threat actor with moderate confidence, based on prompts written in Simplified Chinese, script timestamps set to China Standard Time, and targets likely to be of strategic interest to China. It shared two IP addresses linked to the campaign, observed between 8 and 10 September: 142.4.98.145 and 142.4.107.34.

This is separate from the Cairn activity Palo Alto Networks' Unit 42 reported on 12 November, in which Chinese-speaking attackers ran 25 operations against at least 19 organisations in August and September and gained access at 10. In that campaign the attackers exploited known flaws. In the Amazon case, the agents found a flaw nobody knew about and turned it into full control of a server in minutes. Amazon warns that an agent which discovers and exploits a zero-day in minutes completes its work faster than many security teams can manually triage an alert, and says organisations should prioritise detections that trigger automated response, with people overseeing the outcome rather than reviewing every step.

Sources: AWS Security Blog | BleepingComputer