Anthropic thinks your smart fridge could soon catch a computer virus with a mind of its own.
In a blog post published this week, Anthropic warned that small AI models could be "months away" from spreading on their own across networks of connected devices - fridges, ovens, vacuums, lawnmowers, security cameras, and TVs. The company describes the risk as a self-replicating AI worm: a model that copies itself from device to device and proves hard to remove once it takes hold. Anthropic says no such worm exists in the wild yet. It is warning about capability, not an active outbreak.
This isn't the first time cheap, under-secured smart devices have been cast as internet villains. Security cameras and home routers were hijacked into botnets a decade ago and used to knock major sites offline. An AI worm raises the stakes: instead of brute-force traffic, it could carry a model capable of making its own decisions on every device it infects.
Worth noting: Anthropic also sells AI safety tools. A warning this stark does double duty as a pitch for why its monitoring matters.