“AI Worms”, researchers demonstrate autonomous malware capable of adapting to any online device

Researchers at the University of Toronto have successfully demonstrated the capability of artificial intelligence to power autonomous malware, enabling it to adapt and tailor attacks against a wide range of online devices, including those running Windows, Linux, and IoT systems. This new category of computer worms leverages open-source AI models to autonomously adjust their attack strategies, rendering them highly effective and potentially devastating. The study, published on the arXiv.org preprint server, showcases a proof of concept developed in a controlled environment, highlighting the potential risks and implications of such autonomous malware¹. The ability of these AI-powered worms to evolve and adapt to various systems poses a significant threat, requiring a distinct approach to mitigation and defense. This development shifts the focus from traditional criminal threats to geopolitical state-aligned activities, necessitating a revised playbook for cybersecurity practitioners to counter these emerging threats.