Researchers Build Self-Replicating AI Worm That Operates Entirely on Local, Open-Weight Models

University of Toronto researchers developed a self-replicating AI worm using open-weight LLM to autonomously attack networks without external APIs or human intervention.

• •The worm generates runtime attack logic on a single GPU, tailored to each target, rather than using pre-encoded exploits.
• •In tests, it identified 31.3 vulnerabilities on average and replicated to 62% of the network within seven days.
• •It bypassed its training cutoff by ingesting public CVE advisories and exploited vulnerabilities disclosed after model training.
• •Defenders should segment GPU machines, treat public advisories as weaponization targets, and monitor agent behavioral signals.

This summary was automatically generated by AI based on the original article and may not be fully accurate.