Researchers at Amazon's Threat Intelligence division have documented how Chinese-linked threat actors leveraged Cairn, an open-source autonomous AI penetration-testing framework, to identify what appears to be a zero-day vulnerability during attacks on multiple universities across East Asia.
The investigation commenced on September 10 when Amazon discovered an exposed server running a publicly accessible instance of Cairn. CJ Moses, CISO of Amazon Integrated Security, noted that investigators could access the platform's operational history, including prior prompts. The operator's inputs were written in Simplified Chinese and timestamps pointed to China Standard Time.
Cairn's appeal to attackers lies in its ability to orchestrate multiple AI agents autonomously. When provided with an objective in natural language, the framework can devise and execute strategies without manual intervention at each stage. In this case, the attacker supplied IP addresses and instructed Cairn to "obtain shell access" — essentially requesting the system to identify pathways for gaining operational control.
Moses elaborated: "Based on those inputs, the AI agents planned, executed, and adjusted their approach as they learned, without the operator choosing each next step."
The attackers' efforts included two notable attempts. One sequence involved 148 distinct actions spanning 16 hours, which ultimately failed—including an unsuccessful attempt to exploit the Log4Shell vulnerability. A second effort saw Cairn spend more than six hours attempting to compromise university infrastructure by testing over 30,000 credential combinations derived from 80 legitimate usernames.
The sole successful incursion occurred when the AI agents identified and exploited a path traversal flaw in an open-source medical imaging AI application. At the time of discovery, this vulnerability carried no assigned security advisory and remained unpatched.
Moses stated: "Amazon Threat Intelligence assesses with high confidence that it was a zero-day based on the absence of any corresponding fix, issue, advisory, or prior mention in vulnerability databases for the affected open-source project."
The attackers subsequently deployed the same zero-day to breach a second university. Amazon's team contacted the open-source project maintainers, who released a patch. AWS also collaborated with government partners to alert affected institutions, though the company declined to identify the universities or their location.
What distinguished this campaign was the velocity and relentless pace of the AI agents' operations. Moses cautioned that an agent capable of discovering and weaponising a zero-day within minutes "completes its activity faster than many security operations centers can manually triage an alert." The agents "ran continuously for 16 hours without breaks, sleep, or working-hour constraints."
Artex vs. Cairn
Recent weeks have seen cybersecurity researchers and law enforcement agencies issue multiple alerts regarding AI-driven offensive cyber capabilities deployed by attackers, particularly those originating from China.
The FBI released an advisory one week prior concerning Artex, another AI tool capable of converting any large language model into a sophisticated hacking instrument that can "automate multiple phases of cyber intrusions at machine speed, from reconnaissance through data exfiltration."
Artex has been linked to breaches targeting South Korean and Japanese financial institutions. Microsoft attributed its use by Chinese state-sponsored actors in a recent supply chain compromise of ConnectWise, an IT management software provider. Moses noted last week that Artex appeared in credential theft campaigns against AI models and operations targeting cryptocurrency exchanges.
Google researchers highlighted Cairn on Tuesday in an analysis examining the expanding deployment of autonomous multi-agent offensive frameworks, including Artex, CyberStrikeAI, PentAGI and comparable tools.
Google documented Cairn's use in intrusions against U.S. aerospace companies and confirmed, as AWS discovered, its proficiency at progressing from a target's IP address to achieving operational objectives.
The AWS report references a September study from cybersecurity firm Gambit Security documenting Cairn's deployment by a profit-driven hacking collective targeting online retailers.
Eyal Sela, Director of Threat Intelligence at Gambit Security and lead researcher on that study, told Recorded Future News that Cairn represents a more substantial threat than Artex owing to its general-purpose design.
Sela explained: "While it is good at solving 'red team' or capture-the-flag tasks, it can also solve other problems and the main point is that it has no internal structuring for cybersecurity. When a tool is more generic, it is less predictable because it has less structuring around the way it solves a task."



