AI · 4h ago
Anthropic’s Frontier Red Team says small AI models may be months away from self-replicating on consumer IoT gear, after showing a 26-billion-parameter open-weight model running across 15 off-brand Android TV boxes bought for about $55 each. In a separate lab test, GLM-5.2 completed a full replication cycle across eight emulated Android devices.
The point is not speed. Anthropic split the larger model across multiple tiny boxes and got usable inference, which means the model can be spread across weak devices instead of relying on one server. That matters because the control plane for a future worm could be distributed across cheap hardware, making shutdown harder even if each node is slow and churns out.
For defenders with Android-based IoT fleets or unmanaged consumer devices on internal networks, the shift is toward persistence: the compute can live at the edge and keep reasoning there. The reporting still stops at lab evidence, but it shows the bottleneck is shrinking enough that single-point takedowns may matter less if model capability keeps improving.
6 sources covering this story
Anthropic warns small AI models could self-replicate through IoT devices
Small weight-based replicators" could usher in a new generation of computer worms.
Anthropic warns self-replicating AI could spread to your smart devices
Anthropic is warning that small AI models could soon be able to self-replicate across consumer IoT devices, from Android TV boxes to smart ovens, creating distributed botnets capable of running AI workloads without relying on data centers.
Anthropic Finds Ingredients for Future AI Worms
Anthropic says AI worms may be months away, but its research stops short of demonstrating a practical self-replicating attack.
Anthropic Warns Small AI Models Could Soon Form Self-Propagating IoT Botnets
Anthropic warns small AI models could form IoT botnets; GLM-5.2 propagated Llama3.1-8B to eight nodes in four of five tests.
Anthropic conjures another threat that doesn't exist yet to stoke fears of AI apocalypse
Self-replicating AI botnets could become feasible on IoT hardware within months, claims the Claude maker, if smaller models get smarter
Small AI Models Could Soon Self-Replicate on IoT Devices, Anthropic Warns
Experiments show cheap TV boxes can run capable AI models, and small models may soon be able to spread across such devices on their own.
Part of the PlainSec briefing for 2026-10-14