Anthropic warns of AI ‘worms’ spreading across smart gadgets

Anthropic, a leading AI firm, has cautioned that small artificial intelligence models could soon behave like computer worms. These worms would be able to autonomously spread between internet-connected devices, such as smart home appliances and IoT gadgets, without requiring human intervention or powerful data centers. This new form of cybersecurity threat emerges as tech companies increasingly integrate powerful AI into everyday devices.

The research, conducted by Anthropic, draws parallels to a notorious botnet named Kimwolf, which infected over two million devices using hijacked smart TV boxes. Although the AI worm described in the study would likely propagate slowly, similar to a swarm of fruit flies, it would pose a significant challenge to eradicate due to the lack of a central control mechanism.

While Anthropic's estimates are rough, the researchers believe that the increasing prevalence of AI in consumer devices and the shrinking size of AI models necessary to execute such attacks will only exacerbate the issue. Experts warn that the real threat comes from human malicious actors leveraging AI to enhance their hacking capabilities, rather than AI systems developing a mind of their own.