Anthropic Research Warns AI Models Could Autonomously Replicate Across Unsecured Smart Devices by Year-End

Anthropic's Frontier Red Team has published research demonstrating that small AI models could achieve autonomous self-replication across poorly secured smart TVs, thermostats, and other internet-connected household devices within months—potentially creating AI-powered botnets that lack a centralized command-and-control point. The research, published October 12, 2026, defines autonomous replication as a four-step process: identifying vulnerable IoT devices, exploiting access vulnerabilities, deploying the model to run inference on target hardware, and continuing propagation to additional hosts. Anthropic researchers successfully demonstrated this end-to-end on 15 off-brand Android TV boxes costing roughly $55 each, each equipped with 4 GB of RAM.