Anthropic warns AI worms replicate on Android TV boxes

Anthropic warns that small AI models could soon be capable of self-replicating as AI worms on insecure IoT devices such as off-brand Android TV boxes. In controlled experiments, researchers found that 9B models could run inference on various hardware types, and larger 30B models could already complete replication tasks. Security experts advise that certified Google TV devices carry lower risks, while many cheaper Chinese Android TV boxes and some smart TV apps lack adequate protections. Open article →