Muhammad Umer

UrduGuard: Benchmarking and Detecting Urdu-Language Jailbreak and Prompt-Injection Attacks in Multi-Agent LLM Systems

Jailbreak and prompt-injection defenses for large language models (LLMs) are trained and evaluated almost exclusively on English text, and recent work has shown that this leaves models vulnerable to the same attacks once translated into low-resource languages. This gap is largely...