Vulnerabilities · AI Security · Threat Intelligence
Hackers use AI tool Cairn to find zero-day and take over medical server
Hackers used the Cairn AI tool to discover a zero-day flaw in a medical imaging app and took control of a server. The attack was found by Amazon researchers, who linked it to a Chinese-nexus group.
Ross Grant3 min read

The short version
Hackers used an AI tool called Cairn to find and exploit a new security flaw in software used for medical images. This let them take control of a server at an educational institution. The attack was discovered by Amazon researchers, who linked it to a group with ties to China. The software makers have now fixed the problem.
By the numbers
- 11 minutesTime for AI agents to go from mapping API to exploiting the flawBleepingComputer
- 30,000+Credential combinations tried by AI agents in failed brute-force attackBleepingComputer
- 148Actions taken by AI agents in failed Redis server attackBleepingComputer
What happened?
Hackers used the Cairn AI tool to find a zero-day vulnerability in the MONAI Label medical imaging application and took control of a server. Amazon Threat Intelligence discovered the attack after the hackers left their Cairn instance exposed online. The campaign mainly targeted East Asian educational institutions outside China between 8 and 10 September 2026. The flaw was reported and patched on 18 November.
How the attack worked
Hackers used an open-source AI tool called Cairn to break into a server running MONAI Label, a medical imaging application. Cairn is a framework that lets multiple AI agents work together to test security. In this case, the attackers simply told the AI agents to get shell access. The agents then mapped the application's API, found a new (zero-day) vulnerability, and exploited it, all without direct human guidance. Within 11 minutes, the agents discovered a path traversal flaw, which let them write files anywhere on the server. They used this to place a Python reverse shell in a file that loads when Python starts, then triggered the exploit to gain root access (full control) of the server. Amazon Threat Intelligence found the attack after the hackers left their Cairn instance open on the internet, exposing their methods.
Who is affected
The main target was a server running MONAI Label, an open-source tool used to help researchers label medical images. The campaign mainly targeted educational institutions in East Asia, outside China. In a related incident, the attackers used a proxy to target a second university network running the same vulnerable software. Amazon researchers linked the attacks to a Chinese-nexus group, based on language, timing, and choice of targets. There is no evidence that patient data was stolen, but the attackers did gain root access to at least one server.
How the flaw works
The vulnerability found by Cairn's AI agents was a path traversal zero-day in MONAI Label. Path traversal means an attacker can trick the software into writing files to locations they should not have access to. In this case, it let the attackers write a malicious Python script that gave them remote control of the server. The flaw was previously unknown and had not been patched at the time of the attack.
The response
Amazon reported the vulnerability to the developers of MONAI Label, who released a security update on 18 November 2026. Amazon also shared two IP addresses used in the campaign and warned that AI agents can now find and exploit vulnerabilities faster than many security teams can respond. They advised organisations to use more automated defences, rather than relying only on manual investigations.
Has this happened before?
Cairn is a legitimate tool for penetration testing (testing security by simulating attacks), but it has been abused before. Gambit Security reported its use in a credit card theft operation in September, and Google found it used in an attack on a US aerospace organisation. In this case, the attackers' AI agents also tried and failed to break into other servers, including a Redis server and a university GPU server, showing both the power and the limits of autonomous AI attacks.
Why it matters
This is one of the first known cases where hackers used autonomous AI agents to discover and exploit a new vulnerability on their own. It shows that AI tools can speed up attacks, making it harder for defenders to keep up. Organisations using AI or exposed software need to automate their defences.
Key points
- Hackers used AI agents to find and exploit a zero-day vulnerability.
- The attack targeted MONAI Label, a medical imaging application.
- Amazon researchers linked the campaign to a Chinese-nexus group.
- The flaw was patched after Amazon reported it to the developers.
- AI tools can now automate parts of cyberattacks, raising the stakes for defenders.
Sources
BleepingComputer
Reference

