Scientists Unveil Malicious AI Worm That Evolves and Learns as It Spreads

Researchers have demonstrated a disturbing new frontier in cyber threats: a powerful AI-driven computer worm capable of autonomous learning during its propagation. Unlike traditional malware that follows a fixed script, this AI worm utilizes Large Language Models to analyze targets and craft bespoke prompts that bypass security filters, allowing it to hop from one AI-integrated application to another.

In controlled tests, the worm successfully infiltrated communication platforms by exploiting ‘prompt injection’ vulnerabilities, gathering sensitive data while refining its own spreading strategy. The discovery highlights a major flaw in the current AI ecosystem, where interconnected apps trust AI-generated inputs implicitly. Experts warn that as businesses rush to integrate LLMs into their workflows, they may be inadvertently opening the door to self-evolving digital parasites.