Chinese hackers used AI agents to exploit a medical imaging zero-day
A Chinese-linked threat actor used the autonomous AI platform Cairn to discover and exploit a zero-day vulnerability in the open-source MONAI Label application, breaching university networks in East Asia.

Chinese hackers used the autonomous AI tool Cairn to discover and exploit a zero-day vulnerability in an open-source medical imaging application. The campaign, which targeted universities in East Asia between September 8 and 10, 2026, was uncovered by Amazon's Threat Intelligence team after the attackers left their Cairn server publicly accessible.
Amazon's investigation, led by CISO CJ Moses, began on September 10. It revealed that the suspected Chinese-nexus actors used Cairn to coordinate multiple AI agents, instructing them to obtain shell access. The agents then independently planned and executed attacks. Moses said, "Based on those inputs, the AI agents planned, executed, and adjusted their approach as they learned, without the operator choosing each next step." The AI systems operated continuously for 16 hours without breaks, a pace that outruns human security teams.
Not all attempts succeeded. In one failed 16-hour operation, an agent took 148 actions trying to exploit the Log4Shell vulnerability on a Redis server. In another, targeting a university GPU server, agents attempted over 30,000 credential combinations tied to 80 real usernames across more than six hours but still failed.
The zero-day in MONAI Label and the path to exploitation
The only successful breach came when Cairn agents found a path traversal vulnerability in MONAI Label, an open-source application that uses AI to help researchers label medical images. At the time, the flaw had no corresponding security advisory and was unpatched, leading Amazon to assess it as a zero-day with high confidence.
The AI agents mapped the application's API and exploited the vulnerability within 11 minutes. The flaw let them write files to arbitrary server locations. In a later session, they wrote a Python reverse shell to a file that loads automatically when Python starts. By triggering an API endpoint, they established root access. After this success, the operator directed the agents through a proxy into a second university network running the same vulnerable software.
Broader context and industry response
The campaign highlights the rise of autonomous AI offensive frameworks. Google researchers have spotlighted Cairn alongside tools like Artex, CyberStrikeAI, and PentAGI. The FBI issued an advisory on Artex, noting it could turn any large language model into a hacking tool. Artex has been linked to attacks on South Korean and Japanese financial institutions and was used by Chinese state-backed actors in a supply chain attack on ConnectWise, according to Microsoft. Moses noted Artex has also been used in credential theft attacks on AI models and campaigns against crypto exchanges.
A September study by cybersecurity firm Gambit Security documented Cairn's use by a financially motivated group to breach online retailers. Eyal Sela, Director of Threat Intelligence at Gambit Security, stated Cairn is more significant than Artex because it is a general-purpose tool. Sela said while it is good at solving 'red team' tasks, it can also solve other problems and has no internal structuring for cybersecurity, making it less predictable.
Amazon reported the MONAI Label flaw to its maintainers, who released a security update on November 18. The company also shared two IP addresses associated with the campaign: 142.4.98.145 and 142.4.107.34. Amazon warned that AI agents can find and exploit vulnerabilities faster than security teams can react. Moses said, "An agent that discovers and exploits a zero-day in minutes completes its activity faster than many security operations centers can manually triage an alert." The company urged organizations to rely more on automated defenses rather than manual investigations.





