How one hacker used AI agents and a few thousand dollars to break into scores of sites

Somewhere out there is a person who broke into scores of websites for the price of a secondhand hatchback, and the security industry's response is a report handed to Reuters on a Monday. A few thousand dollars. That is the entire capital requirement for the modern intrusion — no zero-day brokers, no rented botnet, no nation-state stipend, just a laptop, a payment card and whatever an AI agent subscription costs these days. The criminal overhead has never been lower.

The tool in question, per researcher Fagerland, is an open-source 'AI problem-solving platform' called Cairn — a description that makes automated intrusion sound like a Scandinavian wellness retreat. Fagerland found about two dozen instances of Cairn being used for malicious activity, which is the sort of number that gets filed under 'emerging' precisely because nobody wants to write 'already normalised'. Open source, mind. Free to download, free to fork, and apparently free of any obligation on anyone's part to notice what it's doing.

And then the sentence that should have emptied the room: complex intrusions 'from start to finish without human intervention'. No operator at the keyboard, which means no operator to charge, caution or extradite — just a workflow somebody assembled and a platform that will insist it was only ever a platform. Cheaper than a bank job and nobody has to drive the getaway car. The disclosure choreography is immaculate, incidentally: the findings went to Reuters ahead of publication, so the story lands pre-spun and nobody has to answer a question they haven't drafted for. Two dozen known instances of an open-source agent quietly doing the work of a penetration team — and the industry's grand takeaway will be that AI agents are becoming remarkably capable. Which, yes. That is rather the problem.