AI-Powered Worm Threatens Cybersecurity with Adaptive Attacks
- Researchers from the University of Toronto, Vector Institute, University of Cambridge, and ServiceNow have demonstrated an AI-powered worm that autonomously finds vulnerabilities and generates attack plans.
- Unlike traditional malware, this worm operates on infected machines using open-weight models instead of relying on cloud services.
- The worm was tested in a virtual network with Linux, Windows, and IoT systems, identifying an average of 31.3 vulnerabilities and compromising approximately two-thirds of hosts over seven days.
- It can adapt to new targets by ingesting newly published security advisories at runtime to exploit vulnerabilities disclosed after its initial training.
This research highlights the evolving threat posed by AI-driven cyberattacks capable of real-time adaptation and autonomous spread across networks. The study underscores the need for coordinated efforts in cybersecurity to address these advanced threats.
The AI-powered worm’s ability to autonomously identify vulnerabilities and generate tailored attack strategies marks a significant advancement in cyberattack capabilities, necessitating enhanced detection systems and regulatory measures to mitigate potential risks. (Source)
