As of: 2026-10-13 07:38 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/p/x-benjaminmmurphy-2109734358282125537/

# Ben Murphy: 'we are going to quickly need to adapt to a world with omnipresent swarms'

Ben Murphy (@benjaminmmurphy), X, 2026-10-12.
Source: https://x.com/benjaminmmurphy/status/2109734358282125537

## Why it matters

Policy reaction to the Anthropic study from a researcher whose profile lists RAND CAST, IFP and FAI.

## Summary

Posted Oct 12, 2026 (19:54 UTC). Ben Murphy quotes @scaling01's post: 'If small models are capable of self-replication on consumer-grade hardware, such swarms will inevitably exist out in the world. A developer will launch such a model for fun; a misaligned large model will do so intentionally; or criminal groups will build organic botnets based on model-led exploitation.' He argues that 'we need the capability to monitor and shut down these models where necessary, rapidly roll out software patches to targeted hardware, and understand ... what the agents' aims are'.

_Metrics as of 2026-10-13 via api.fxtwitter.com: 6,736 views, 72 likes, 22 reposts._

## Archived text

> If small models are capable of self-replication on consumer-grade hardware, such swarms will inevitably exist out in the world. A developer will launch such a model for fun; a misaligned large model will do so intentionally; or criminal groups will build organic botnets based on model-led exploitation.
> 
> While control and monitoring remain incredibly important for risks caused by lab-hosted agents, we are going to quickly need to adapt to a world with omnipresent swarms. This means we need the capability to monitor and shut down these models where necessary, rapidly roll out software patches to targeted hardware, and understand—from external signals or from the swarm itself—what the agents' aims are.
> 
> > Quoting @scaling01: ANTHROPIC: "small AI models may be months away from being capable enough to self-replicate on commercial IoT devices"
> > 
> > "In our evals, a 9B model got inference running on 8 of the 9 hardware types we tested"

Archived 2026-10-13 via fxtwitter (unofficial).
Counts: 6,836 views, 72 likes, 22 reposts, 7 replies (at fetch time)

## Cited in

- 2026-10-12: [Anthropic's Frontier Red Team](https://postcutoff.com/e/2026-10-12-anthropic-self-replicating-ai-iot-devices/)