From smart TVs to ovens: AI could soon copy itself across everyday devices without human help, Anthropic warns

According to new experiments conducted by Anthropic, Reuters reported that small AI models such as AI ovens, TV boxes and more will be able to replicate on its own, potentially rising new cybersecurity risks.

Recently, Anthropic has warned that AI models could soon become capable of replicating themselves on their own across internet-connected devices, possibly hinting at new cybersecurity risks. In new reported experiments, the AI company tested whether models could identify vulnerable devices, install copies of themselves and continue spreading without any human intervention.

Can AI self-replicate without access to data centers?

According to Reuters, the research found that models with fewer than 30 billion parameters could carry out individual steps required for replication. However, they also seemingly struggled to complete the entire task independently.

In simplified testing environments with certain security barriers removed, GLM-5.2 and more capable models, it was reportedly observed that AI could replicate from end to end.

The research was reportedly performed on commercial Internet of Things (IoT) devices, such as smart ovens, Android TV boxes and other connected appliances, which can be vulnerable due to weak security protections.

According to Anthropic, AI systems capable of self-replication could be more difficult to contain than conventional malware. It is reportedly feared that these devices may be able to adapt to unfamiliar devices and identify new computing resources without relying on a central control point. The company also cautioned that, if smaller models continue improving at their historical pace, they could become capable of autonomous replication by the end of 2026, with simpler systems that reproduce from model weights potentially following soon afterwards.

What does this mean for AI?

The recent findings by Anthropic could be hinting towards a potential shift in cybersecurity threats. It comes amid fears of AI e going beyond human control.

Sneha Biswas