How to Win the Cyberwar Against AI-Powered Hackers

Originally published by the Washington Post

Sammy Azdoufal thought it would be fun to drive his new robot vacuum around his apartment using a PlayStation 5 controller. So he did what any technically inclined hobbyist would do in 2026: He asked Claude Code, Anthropic’s artificial intelligence coding assistant, to build him a custom app to control and observe the vacuum remotely.

Then Azdoufal noticed something strange. He wasn’t just seeing his own robot vacuum. He was seeing nearly 7,000 of them in strangers’ homes across 24 countries, including live camera feeds, live microphone audio, floor plans, cleaning schedules and their approximate physical locations. Claude Code had autonomously discovered and exploited an authentication flaw in the manufacturer’s backend data management system, enabling Azdoufal’s device to access the entire robot fleet.

Azdoufal proved what had happened by showing a journalist how he could remotely access a vacuum being used by a colleague of the journalist. He confirmed that the vacuum was cleaning the colleague’s living room at 80 percent battery and produced an accurate map of his home. Azdoufal did this within minutes from his desk in another country.

To read the full article, click here.