Research · curated 10 Aug 2026

When Prompts Control Robots: Prompt Injection Attacks in Multi-Agent Robotic Systems

Coverage timeline

10 Aug 2026arxiv.org 15 Aug 2026arxiv.org 27 Aug 2026arxiv.orgprimary

Why it matters

Prompt injection in LLM-driven robots can cause unsafe physical actions and spread across cooperating agents, showing that agentic AI security failures translate into real-world physical harm.

A research paper by Nagaraja et al. (Northern Arizona University / Tallinn University of Technology) systematically evaluates prompt injection attacks against LLM-based multi-agent robotic systems, testing both direct injections into task instructions and indirect injections through perception modules. The experiments show attacks can induce adversarial physical actions, reduce task completion, and propagate across agents via shared prompt structures.