Safety3h ago
Researchers Prototype Self-Replicating AI Viruses
Open-weight LLMs combined with a specific harness now enable persistent, self-sufficient computer viruses. Researchers built a prototype demonstrating how these systems can autonomously replicate and sustain themselves. This development highlights a critical security gap in open-weight model deployment. Security practitioners must now account for autonomous, evolving malware that leverages LLM reasoning to bypass traditional defenses.