4 min read
Research Researchers from Toronto, Cambridge, and ServiceNow demonstrated an AI worm that ingests public vulnerability advisories at runtime and synthesises working exploits for CVEs it was never trained on, successfully compromising targets across a simulated network.