Anthropic: AI models Will Soon Spread Across Your Home Devices Without Control

Artificial intelligence models could be months away from developing capabilities to replicate themselves across internet-connected household devices, Anthropic now says, warning that such systems could be difficult to shut down and pose growing cybersecurity risks.

The AI company warns that small models could potentially infiltrate devices such as smart ovens and Android TV boxes, using their computing power to run themselves and spread to other machines without relying on expensive data centres.

The AI model would be “challenging to shut down,” Anthropic says in a blog post, “since there’s no centralized point to shut off.”

Anthropic notes that it had not yet observed a “fully rogue replicator,” but that AI agents were increasingly capable of completing individual steps needed for replication and passing adaptations to future copies of themselves.

The findings add to growing concerns among researchers and technology companies about the security implications of increasingly capable AI systems, including their potential to automate cyberattacks, exploit vulnerable devices and operate beyond effective human oversight.

Anthropic says it conducted experiments to assess whether AI models could replicate across networks of devices. One test involved running Google’s Gemma4-26B-A4B model across 15 inexpensive, off-brand Android TV boxes.

The company notes that such devices often have weak security, potentially making them vulnerable to infiltration. Models too large to run on a single device could be divided into smaller components, or “shards”, distributed across multiple machines.

“In practice, an AI worm wouldn’t necessarily consist of just one model or strategy. A swarm of different models deployed on different types of hardware using a variety of replication schemes might represent a more realistic ‘worst case scenario’ attack,” Anthropic warns.

The company says it remains unclear how autonomous AI replication might emerge, but outlined several possible scenarios, ranging from slow-moving networks of compromised devices to more powerful systems capable of breaching data centres or obtaining computing resources through fraud.

It intimates that future AI agents could operate across a wide range of hardware, including devices outside human control.

“As a prediction that goes beyond our experiments, we expect a future awash with agents running on all types of hardware, including some that aren’t under human control,” Anthropic explains. “The rise of slow replicators would also be a bellwether for more aggressive, faster ones that can breach more hardened machines or be effective API key thieves.”

The research highlights a potential weakness in the expanding network of connected consumer devices, many of which have limited computing resources and inconsistent security protections. A compromised device could provide a foothold for an AI system to spread across a network, although successful replication would depend on vulnerabilities, available computing power and the ability to evade detection.

The prospect also complicates cybersecurity efforts because distributed systems may lack a single server or central control point that operators can disable. Defenders could instead need to identify and isolate compromised devices across multiple networks.

Stay ahead of the Stories shaping our world. Subscribe to Impact Newswire and join our 
WhatsApp Channel for updates on global tech, business, and innovation—all in one place.

Dive deeper into the future with the Cause Effect 4.0 Podcast, where we explore the ideas, trends, and technologies driving the global AI conversation.

Got a story to share? Contact Us to reach a global audience with Impact Newswire.