AI worms could spread across smart gadgets, Anthropic says

Small artificial intelligence models may be only months away from being able to behave like computer worms, copying themselves from one internet-connected gadget to another without human help, Anthropic said.

The AI company said the risk could open a new cybersecurity front as tech firms build more AI directly into phones, laptops, cars and household devices. A computer worm is malicious software that spreads between machines on its own.

Anthropic researchers William Zhou and Daniel Freeman said an AI worm would not need powerful data centres to operate. They said it could instead run on networks of cheap, poorly secured consumer devices such as smart TV boxes and kitchen appliances.

The researchers based their estimate on Kimwolf, a botnet that infected more than two million devices at its peak, mostly off-brand Android TV boxes. They said an AI worm of this kind would likely spread slowly, taking days to copy itself, but would be hard to eliminate because there would be no central switch to turn off.

The researchers said the estimate was rough because it was based on one type of gadget and one past botnet. They said the risk could grow as everyday gadgets gain computing power and the AI models needed for such behaviour become smaller.