AI Hacking Tool Cairn Finds Apparent Zero-Day and Breaches Universities, Amazon Says
Amazon's Threat Intelligence team reports that attackers used an AI tool called Cairn to breach universities in East Asia. The investigation began on September 10, when Amazon found a publicly accessible server hosting an instance of Cairn, an open-source framework that coordinates multiple AI agents to carry out penetration-testing tasks. The server exposed the operator's prompt history, which was written in Simplified Chinese with timestamps set to China Standard Time.
The attacker gave Cairn IP addresses and asked it to obtain shell access. According to CJ Moses, CISO of Amazon Integrated Security, the agents then planned, executed and adjusted their approach without the operator choosing each step. Most attempts failed, including a 148-action effort over 16 hours that involved a failed Log4Shell attempt, and a six-hour run testing more than 30,000 credential combinations against 80 real usernames.
One attempt succeeded. The agents found a path traversal flaw in an open-source medical imaging AI application and exploited it. Amazon assesses with high confidence that it was a zero-day, as no fix, advisory or prior mention existed. The attacker reused the flaw to breach a second university. Amazon contacted the maintainers, who fixed the bug, and AWS worked with a government partner to notify the affected organisations. AWS did not name the universities or the country.
Moses stressed the speed of the campaign. An agent that finds and exploits a flaw in minutes can finish before many security operations centres have triaged an alert, and these agents ran for 16 hours without breaks. Google has also reported Cairn in intrusions at U.S. aerospace companies, and Gambit Security documented its use against online retailers. Gambit's Eyal Sela said Cairn is a general-purpose tool with no internal cybersecurity structuring, which makes it less predictable.
Why it matters
The campaign challenges the assumption that defenders have time to triage alerts and patch after a flaw is disclosed, since an agent can find and exploit an unreported bug in minutes. Organisations running niche open-source applications, such as medical or research tools, are exposed because they may have no advisory to patch against. Because Cairn is general-purpose and also seen used by financially motivated groups, small businesses outside the university sector cannot treat this as a targeted espionage issue.
What to do
- Inventory internet-facing open-source applications and remove public access to any that do not need it.
- Check for Log4Shell exposure, since Cairn attempted it, and patch any vulnerable systems.
- Review authentication logs for large volumes of failed logins across many usernames.
- Set up out-of-hours alerting so suspicious activity is reviewed without waiting for working hours.