Anthropic warned that artificial-intelligence “worms” capable of self-replication may soon be able to spread across internet-connected home devices, exploiting their comparatively weak defenses. The company said stronger models already show “end-to-end replication,” a capability it expects smaller systems could reach within months, complicating shutdown efforts because there is no centralized kill switch. Security firms have flagged AI-driven malware’s ability to move quickly and evade detection, raising concerns about swarming agent attacks on TVs, ovens, cameras and other IoT hardware. The researchers urged hardening of systems and preparation for persistent threats, noting possible attack patterns could range from “digital slime mold” behavior to models attempting to breach data centers.
Related article:
