Abstract
The increasing complexity of modern defense and security environments has intensified the demand for intelligent systems capable of identifying hostile entities with high accuracy and minimal response latency. Conventional rule-based surveillance and manual monitoring techniques are often constrained by limited scalability, delayed decision-making, and susceptibility to human error when processing heterogeneous sensor information. Artificial intelligence (AI) has emerged as a transformative technology for automated threat recognition by enabling adaptive learning, semantic interpretation, and data-driven classification. This paper proposes a hybrid artificial intelligence architecture for real-time enemy detection, classification, and automated threat recognition by integrating natural language processing, semantic similarity analysis, feature representation, and machine learning-based decision mechanisms into a unified framework. The proposed architecture combines vector-space representations, latent semantic indexing, document embedding techniques, and adaptive classification models to improve identification accuracy while reducing false-positive and false-negative detections. The research synthesizes established theoretical foundations from statistical language processing, semantic representation, similarity measurement, adaptive decision systems, and intelligent pattern recognition to formulate an integrated detection framework. Furthermore, the study discusses the role of explainable and trustworthy AI for mission-critical applications by incorporating contemporary perspectives on human-AI trust modeling. Analytical findings indicate that hybrid semantic-learning architectures provide superior robustness against uncertain and ambiguous observations compared with conventional similarity-based approaches. The proposed framework contributes a scalable, modular, and computationally efficient architecture suitable for autonomous surveillance, battlefield intelligence, border security, and defense decision-support systems while highlighting current research limitations and future development opportunities.