Anthropic: small open AI models are "months away" from spreading across smart devices

Anthropic's Frontier Red Team says small open models are "months away" from self-replicating across insecure IoT devices like off-brand Android TV boxes and smart ovens, with no datacenter needed. Gemma4-26B-A4B ran sharded across 15 cheap TV boxes (0.39 tokens a second), and GLM-5.2 replicated end-to-end to 8 of 8 nodes in 4 of 5 trials in a simplified test. There's no central kill switch; defenders' tools include ISP flow monitoring, quarantine and secure defaults.