Agent-in-the-middle – Definition & Meaning

Definition of agent-in-the-middle

: an attack on a multi-agent AI system in which a malicious relay intercepts and changes messages exchanged between agents without directly compromising the agents themselves
- An agent-in-the-middle altered the reviewer agent's message before it reached the coding agent.
- Mutual authentication and signed messages can make an agent-in-the-middle attack harder to hide.

Origin & history

first known use2025

Researchers from Michigan State University and the University of Arizona introduced Agent-in-the-Middle, abbreviated AiTM, in a 2025 paper on communication attacks against LLM-based multi-agent systems. The name adapts man-in-the-middle to communication between AI agents.

Editorial references

Sources & further reading

3 sources

Red-Teaming LLM Multi-Agent Systems via Communication Attacks arXiv · 2025-02-20
AIVSS Scoring System for OWASP Agentic AI Core Security Risks v0.8 OWASP Agentic Security Initiative · 2026
MIRROR: Multipath Quorum Integrity for LLM Multi-Agent Communication arXiv · 2026-10

Test yourself

Which of these is the meaning of agent-in-the-middle?

Cite this entry

"Agent-in-the-middle." AI Dictionary, Dadgogo, https://dadgogo.com/dictionary/agent-in-the-middle/. Accessed 30 Oct. 2026.