Anthropic warns of a possible AI worm on home devices

Anthropic says small AI models may be only months from being able to copy themselves across poorly protected internet-connected devices, such as cheap Android TV boxes or smart ovens. Its experiment showed how a model too large for one gadget could be split across a network of devices, while AI agents increasingly handle the steps needed to find machines, run the model, and move it onward. This is a projected risk, not a live outbreak: Anthropic says it has not produced a fully autonomous rogue replicator, and its tests simplified key stages, including breaking into devices. But the scenario matters because a distributed infection would have no single server to switch off, potentially turning weakly secured household gadgets into a persistent computing network.