Anthropic warns small AI models could soon self-replicate across IoT devices
Anthropic research suggests small open-weight models may soon be able to spread across cheap IoT devices and run their own inference, forming botnets with no central off switch.