Advancing Early Biorisk Assessment, Emerging Infectious Disease Forecasting, and Rapid Decision-Making Through a Novel AI-Based One Health Platform

Emerging infectious diseases continue to threaten global public health, veterinary medicine, agriculture, environmental security, and economic stability. Although artificial intelligence has significantly improved epidemic forecasting, most existing approaches remain reactive, focusing on predicting disease spread after outbreaks have already been detected. Consequently, current surveillance systems provide limited capability for identifying conditions associated with pathogen emergence and supporting proactive biological risk assessment.Unlike conventional forecasting systems that primarily respond to already established outbreaks, the proposed framework aims to identify upstream biological, environmental, and epidemiological conditions associated with pathogen emergence before sustained epidemic transmission occurs. By integrating heterogeneous One Health information, the framework aims to identify early biological risk signals and support preventive interventions prior to large-scale epidemic development.This paper presents Pandemic Radar AI, a prototype AI-based One Health framework designed to support early biorisk assessment, pandemic forecasting, and evidence-based decision-making through the integration of heterogeneous biological, environmental, and epidemiological information. Unlike conventional forecasting models, the proposed framework combines epidemiological surveillance, virological, veterinary, environmental, climatic, demographic, and geospatial data within a unified multimodal analytical architecture to identify conditions associated with pathogen emergence, identify biological risk hotspots, forecast epidemic dynamics, and generate decision-support indicators for public health authorities.The prototype implements a scalable modular architecture integrating machine learning, artificial neural networks, cellular automata, multimodal data fusion, semantic data integration, feature engineering, uncertainty-aware risk assessment, explainable artificial intelligence (XAI), and AI-assisted decision support. Artificial intelligence methods were selected because epidemic forecasting is characterized by high uncertainty, nonlinear dynamics, and heterogeneous multimodal data. Complementary AI approaches therefore enable more robust modelling of complex epidemiological processes than any individual algorithm alone. To demonstrate technical feasibility, a pilot forecasting module was developed and evaluated using publicly available SARS-CoV-2 epidemic data as a representative case study. The present study focuses on the conceptual framework and prototype implementation rather than on a complete computational description of all analytical modules. The proposed architecture is intended as a scalable foundation for future development and validation using multiple infectious diseases and heterogeneous One Health data sources.Preliminary experiments demonstrated reliable short-term forecasting using SARS-CoV-2 epidemic data. Forecasting performance gradually decreased with increasing prediction horizon, while useful predictive capability was maintained for horizons of up to 14 days. Maintaining useful predictive performance over a 14-day forecasting horizon is particularly important for public health preparedness because it provides additional time for surveillance, resource allocation, risk communication, and implementation of preventive interventions.The proposed framework extends conventional epidemic forecasting by introducing an integrated approach that links biological risk assessment, outbreak forecasting, and decision support within a single AI ecosystem. The framework is designed to support explainable and uncertainty-aware AI-assisted decision making by integrating heterogeneous One Health information into a unified analytical environment suitable for continuous biosurveillance. The presented prototype establishes a foundation for next-generation AI-supported biosurveillance systems capable of continuously integrating multimodal One Health information to identify biological risk conditions, identify conditions associated with pathogen emergence, extend practical epidemic forecasting horizons, and support preventive public health interventions before widespread epidemic transmission occurs. The proposed framework is designed as an evolving research platform that can be progressively expanded and validated across diverse emerging pathogens, geographical regions, and multimodal One Health data sources. The modular architecture also provides a foundation for future integration of geospatial Earth observation data and cybersecurity components to strengthen biosurveillance resilience, environmental monitoring, protection of critical public health infrastructures, and preparedness for emerging biological threats.