(netskope.com) AI Worms and IoT: Revisiting Historical Lessons in Modern Threat Landscapes In brief - This article discusses the potential for AI-driven self-replicating worms to target IoT devices, critiquing recent experiments by Anthropic as unrealistic and overly speculative. Technically - This article analyzes two Anthropic experiments: one involving the sharding of a small model across 15 Android TV boxes for on-device inference, and another deploying an open-weight model to eight GCP nodes running Android emulators. The author argues these tests lacked realism because they bypassed critical attack chain phases—such as network enumeration, authentication, and privilege escalation—and relied on scripts rather than autonomous agency. The author concludes that existing network security controls, behavior monitoring, and access management are sufficient to mitigate these theoretical AI-agent threats. Source: https://www.netskope.com/blog/ai-tries-and-fails-to-worm-its-way-into-your-iot-gadgets #Cybersecurity