AI is making coding more accessible, but it doesn’t always understand the real-world consequences, as one DJI vacuum owner discovered.
Sammy Azdoufal used Anthropic’s Claude Code to hack his DJI Romo vacuum to work with a PS5 controller. Claude successfully reverse-engineered the communication protocol, but the resulting tool accidentally gave access to other users’ vacuums and power stations, including floor plans and camera feeds, due to weak device authentication.
DJI has since fixed the main vulnerability, though some risks, like overriding PINs for camera access, remain. The case highlights how AI-generated code can have unexpected real-world consequences when it interacts with connected devices.
Source: Android Authority
