This research evaluates how effectively large language models can detect hate speech across different languages and cultural contexts, addressing a critical security challenge in content moderation.

• Systematically tests LLM performance across multilingual datasets from diverse geographic regions
• Examines model robustness against adversarial attacks in hate speech detection
• Highlights the importance of cultural and contextual factors in accurate content moderation
• Provides insights for developing more geographically-aware AI systems for security applications

This work offers significant value for security professionals building more effective and culturally-sensitive content moderation systems that can adapt to regional linguistic variations and contextual nuances.

Evaluation of Hate Speech Detection Using Large Language Models and Geographical Contextualization