Abstract: Large Language Models (LLMs) are increasingly integrated into various infrastructure and interactive applications. However, their inherent linguistic flexibility introduces security ...