Amazon says Cairn AI agents found a zero-day in MONAI Label

Amazon says a suspected Chinese-linked operator told Cairn's agents to get shell access at East Asian universities, and they found and exploited an unpatched flaw on their own.

SecurityPublished
What happened
Amazon Threat Intelligence says attackers used the open-source AI pentesting framework Cairn to find and exploit a zero-day at East Asian universities.
What’s new
BleepingComputer reports that the flaw was a path traversal bug in MONAI Label, which its maintainers fixed on 18 November.
Why it matters
Amazon warns that autonomous agents can find and exploit flaws faster than security teams can react to alerts.

Amazon Threat Intelligence says an attacker used Cairn, an open-source framework that coordinates multiple AI agents for penetration testing, to find and exploit a previously unknown vulnerability while attacking universities in East Asia. Amazon spotted the operation on 10 September, after finding a Cairn instance left publicly exposed. CJ Moses, CISO of Amazon Integrated Security, told The Record that this exposure let investigators read the operator's prompts and history. BleepingComputer reports that the activity ran from 8 to 10 September 2026.

The operator supplied IP addresses and told the agents to get shell access, then left them to plan their own approach. Many attempts failed. One effort ran for 16 hours and 148 actions without success. Another tried more than 30,000 credential combinations over six hours and also failed.

The success came in MONAI Label, an open-source tool for labelling medical images. According to BleepingComputer, Amazon says the agents found a path traversal flaw there and went from mapping the application's API to exploiting the flaw within 11 minutes. They later planted a Python reverse shell to gain root access. The attacker then used the same flaw against a second university. Amazon reported the bug, and the MONAI Label maintainers released a fix on 18 November. AWS declined to name the universities or their country.

Amazon links the campaign to a Chinese-nexus actor with moderate confidence, based on the prompt language, timestamps and choice of targets. Moses warned that such an agent "completes its activity faster than many security operations centers can manually triage an alert." Eyal Sela of Gambit Security, which documented Cairn abuse in September, told The Record he considers Cairn more significant than the AI tool Artex because it is general-purpose.