Scientists Find Way to Supercharge Dangerous Computer ‘Worms’ With A.I.

June 3, 2026 - 04:06

A team from the University of Toronto has demonstrated a troubling new capability: using artificial intelligence to build computer worms that can automatically exploit any known security flaw. The researchers showed that a large language model, when combined with existing malicious code, can be turned into a self-replicating program that hunts for vulnerabilities and spreads without human guidance.

The project, which the team calls "Morris II" in a nod to the infamous 1988 worm, represents a significant shift in how cyberattacks might work. Traditional worms need a programmer to write specific code for each target. But this new approach uses an AI model to scan systems, identify weaknesses, and craft the exact exploit needed. In tests, the worm successfully targeted systems running common software and replicated itself across networks.

The implications are stark. Security experts have long worried about automated attacks, but this proof-of-concept shows the gap between theory and reality is closing. The researchers argue that their work is a warning. If AI can be used to supercharge worms, then defensive systems must also evolve. They call for new safeguards in how AI models are trained and deployed, especially when given access to networks or code execution.

While the worm was built in a controlled lab, the team believes similar techniques could be used by real attackers. The study highlights a future where a single AI-driven worm could target every unpatched computer on the internet, turning a slow, manual process into a fast, automated threat.