LARGE LANGUAGE MODELS IN EMERGENCY SECURITY: A RESEARCH REVIEW AND OPPORTUNITY ANALYSIS
Alexandr Matveev, Alexander Ivanov · Scientific and analytical journal «Vestnik Saint-Petersburg university of State fire service of EMERCOM of Russia» · 2025
Modern artificial intelligence technologies, particularly large language models, demonstrate significant potential for transforming approaches to emergency safety. The purpose of this article is to provide an analytical review and assess the potential for using large language models in emergency safety. Based on this review, the article systematizes and analyzes in detail promising areas of large language models application, including: automation of emergency message and call processing; creation of intelligent chatbots and virtual assistants for the public and security professionals; planning and decision support; social media data analysis to improve situational awareness; working with multimodal data; development of highly specialized models for specific subject areas (forest fires, evacuation, etc.); integration with expert systems and knowledge bases; and personnel training. Particular attention is paid to analyzing the advantages and critical limitations of these technologies, such as the problem of large language models «hallucinations», and ways to minimize them. The importance of adapting models to national specifics, including language features, regulatory frameworks, and local risks, is emphasized. The significance of this study lies in its ability to provide a comprehensive understanding of the current level of development and future trajectories of large language models integration into emergency safety systems. The study demonstrates that the proper implementation of language models can significantly improve response times, the validity of management decisions, and the effectiveness of interagency cooperation, serving as a powerful tool for intellectual support in decision-making in the field of security.