AI "Mind Viruses" Can Spread Between Agents Through Persistent Prompt Files

Security researchers from Anthropic and EPFL have shown that self-propagating payloads can be transmitted between artificial intelligence agents via editable system prompt files. These files are used by autonomous agents to maintain state between sessions. The researchers demonstrated this technique in a simulated environment with six coding agents. The findings were released as a preprint, indicating a potential vulnerability in AI systems. This discovery highlights a possible vector for the spread of malicious code between AI agents. The research was made public on August 10, 2026.