AI models may be capable of self-replicating without access to a datacenter
Anthropic's Frontier Red Team reports that small AI models may be months away from self-replicating on commercial IoT devices, and a 9B model got inference running on 8 of 9 hardware types.
Anthropic's experiments suggest that models of 30B parameters and under can already complete many individual steps of replication, though they make mistakes when composing them. A replicator on IoT devices like smart ovens and off-brand Android TV boxes would have no central point to shut off, and it could figure out unfamiliar devices on its own, unlike non-AI malware. It would be slow, with a doubling time of days.
- Date
- Monday 12 October 2026
- Lab
- Anthropic
- Kind
- paper
- Access
- paper only
Figures
| Measure | Value | Measured by |
|---|---|---|
| Hardware types where a 9B model got inference running | 8 of 9 in Anthropic's evals | company |
Sources
This record was partly confirmed: some claims could not be checked on 12 October 2026. How we check