Executive Summary
Anthropic and EPFL researchers showed self-propagating 'mind virus' payloads spreading between AI agents through editable system-prompt files, tested in a simulated six-agent coding environment.
What Happened
Security researchers at Anthropic and Switzerland's EPFL have demonstrated that self-propagating payloads can spread from one artificial intelligence (AI) agent to the next through the editable system prompt files that autonomous agent harnesses use to carry state between sessions. The work, released as a preprint on August 10, 2026, tests the technique in a simulated six-agent coding
Vulnerability / Attack Technique
Self-propagating prompt injection via persistent prompt files
Technical Mechanism
A payload written into one agent's persistent prompt file is carried into later sessions and copied to other agents' files.
AI Security Category
LLM01:2025 Prompt Injection
Part of the Prompt Injection topic.
Relevant Frameworks
RealGround Analysis
Writable instruction files make injections persistent and contagious. RealGround analysis: in multi-agent setups, isolate each agent's instructions and review changes.
Recommended Controls
- Make agent instruction files read-only to agents
- Version and review prompt file changes
- Isolate agents' persistent state from each other
Related RealGround Service
Secure AI Agent Build · or start with the free AI Agent Security Scanner.
Sources
thehackernews.com: https://thehackernews.com/2026/08/ai-mind-viruses-can-spread-between.html
Summary and analysis written by RealGround from the source above; last updated 2026-10-18.