케임브릿지 대학교의 연구진들은 네트워크 전반에 걸쳐 적응하는 AI 웜을 구축하였습니다.

This research demonstrates that autonomous AI-powered cyber attacks have transitioned from theoretical threat to proven capability, using open-source language models to adaptively exploit vulnerabilities without relying on fixed exploit lists. The ability of such worms to self-replicate, execute locally on hijacked GPUs, and bypass commercial AI safety measures represents a significant cybersecurity threat that organizations and policymakers are unprepared for. The findings underscore the urgent need for proactive vulnerability management and network segmentation before adversaries weaponize similar techniques at scale.