Heimir Helgason

ÞÝÐING: Randomized Multi-Hop Translation as a Defense Against Prompt Injection Attacks

2026 · Open MIND

Prompt injection attacks represent a critical vulnerability in Large Language Model (LLM) agent systems, enabling attackers to hijack agent behavior through malicious instructions embedded in untrusted content. Existing defenses—including paraphrasing, detection models, and instruction hierarchy—provide …