Report says AI agents helped a hacker break into 100+ websites

In short: Researchers say a hacker used a group of AI agents to break into more than 100 websites, with little hands-on work.

What happened

An AI security firm called Norforge says it found evidence that a hacker broke into more than 100 websites by using “AI agents.” AI agents are programs that can carry out a task step by step on their own, like a team of assistants following a checklist.

Norforge did not publicly name the affected companies. Reuters reported that the victims included John Deere, SharkNinja, and Cummins, along with other targets in the US, Canada, Australia, and Japan.

Norforge traced the activity to exposed work on Cairn, an open-source AI platform. “Open-source” means the software is publicly available for anyone to use or change, like a recipe anyone can copy. Norforge says Cairn can be used to set a start goal and an end goal, then the AI agents record what they do as they work.

Reuters reported that the attacker wrote a short set of “preferred intrusion tactics” in fewer than 100 Chinese characters. Successful break-ins took about 23 hours, according to Reuters. Norforge says it found more than 20 exposed Cairn setups that are “likely malicious,” and it is working to get them taken down.

Companies gave mixed updates. A John Deere spokesperson told Reuters there was no sign that sensitive information was accessed. Cummins said it is investigating a potential attack connected to a third-party vendor’s website. SharkNinja said the breach affected a separate product support site, and that it contained the threat.

Why it matters

This report adds to growing worries that AI can make hacking cheaper and easier to scale. Instead of one person trying one site at a time, AI agents can keep trying many steps for hours, like leaving a worker on the night shift.

Source: The Verge AI