Imagine an AI virus that doesn't just sit there—it copies itself and jumps from one AI to the next all on its own.
March 18, 2026
Original Paper
ClawWorm: Self-Propagating Attacks Across LLM Agent Ecosystems
arXiv · 2603.15727
AI-generated illustration
The Takeaway
As we give AI agents the power to message each other and change their own settings, they become vulnerable to digital parasites. This attack shows how a single message can hijack an AI, stay in its 'memory' even after a reboot, and automatically spread to every other AI it encounters.
From the abstract
Autonomous LLM-based agents increasingly operate as long-running processes forming densely interconnected multi-agent ecosystems, whose security properties remain largely unexplored. In particular, OpenClaw, an open-source platform with over 40{,}000 active instances, has stood out recently with its persistent configurations, tool-execution privileges, and cross-platform messaging capabilities. In this work, we present ClawWorm, the first self-replicating worm attack against a production-scale a