AI could soon self-replicate on smart devices, Anthropic warns
What's the story
Anthropic has issued a warning about the potential for AI models to self-replicate on Internet of Things (IoT) devices such as smart fridges and ovens. The company says this development could make it difficult for humans to intervene if things go awry. The warning comes as part of Anthropic's ongoing research into the capabilities and risks associated with advanced AI systems.
Replication risk
AI self-replicator could be hard to shut down
An AI self-replicator is a system that can create copies of itself without human intervention.
Anthropic's blog post highlighted the potential dangers of such technology, noting that a replicator would be hard to shut down as "there's no centralized point to shut off," unlike a data center.
The company also said while this AI swarm may spread slowly, it could take days for them to double their numbers.
Research findings
Anthropic tested its theory on cheap Android TV boxes
Anthropic's research focused on what it calls weight-based replicators, or autonomous worms that run their models directly on compromised hardware.
The company found that home and small business IoT devices could be early targets for these systems due to their internet connectivity and historically limited security protections.
To test this theory, Anthropic ran Gemma4-26B-A4B across 15 affordable Android TV boxes at a rate of 0.39 tokens per second.
Model performance
Small models could reach this capability by late 2026
Anthropic's researchers found that while small models still make mistakes when asked to complete the full replication chain end to end, models of 30 billion parameters and under can already complete many individual steps.
The company also noted that if small models improve at their historical rate, they could match this capability level by late 2026 with small weight-based replicators becoming possible soon after.
Security measures
Anthropic calls for securing IoT devices against potential AI swarms
In light of these findings, Anthropic has stressed the need to secure small IoT devices against potential AI swarms.
The company said, "We should aim to harden our systems to make it difficult for such replicators to spread, but also prepare to deal with a world where they are here to stay."
This statement underscores the importance of proactive security measures in an era where advanced AI technologies pose new challenges.