Tokyo, one of the world’s most densely populated megacities, lies in a seismically active region where the threat of major earthquakes is constant. A key concern during such events is soil liquefaction—where saturated, loose soils temporarily lose strength and behave like liquid due to intense shaking. This phenomenon has caused catastrophic damage in past disasters, including the 1995 Great Hanshin-Awaji, the 2011 Great East Japan, and the 2024 Noto Peninsula earthquakes. Despite this, traditional liquefaction hazard maps for Tokyo have been limited by coarse resolution, typically using 500-meter or larger grid scales and sparse borehole data, failing to capture the city’s complex subsurface conditions, especially in reclaimed coastal areas and river floodplains. To address this gap, a research team led by Professor Shinya Inazumi from Shibaura Institute of Technology has developed a groundbreaking approach using artificial neural networks (ANNs) to generate high-resolution 3D liquefaction hazard maps at a 200-meter grid scale. Their method integrates an extensive dataset of 13,926 borehole records—one of the largest AI applications in geotechnical hazard assessment in Japan—with machine learning to predict both soil type and N-values (a critical measure of soil density and strength) at unsampled locations. The ANN model demonstrated superior accuracy in capturing nonlinear geological patterns, significantly outperforming conventional geostatistical methods and other machine learning models. Using these predictions, the team calculated liquefaction potential indices across Tokyo, producing a hazard map with unprecedented detail. The new map clearly identifies high-risk zones in areas such as Koto Ward, which are particularly vulnerable due to soft, reclaimed soils but often underrepresented on official maps. This level of precision enables more targeted urban planning, better foundation design, and informed prioritization of soil improvement projects. The framework is not only highly accurate but also scalable and adaptable, making it suitable for integration into geographic information systems (GIS) for dynamic visualization and public engagement. It represents a major leap forward in proactive disaster risk management, offering a transferable model for other megacities facing similar geohazards. As Professor Inazumi emphasizes, this work bridges advanced AI with civil engineering, supporting safer, more resilient urban development in earthquake-prone regions. Industry experts commend the study as a pivotal advancement in geotechnical risk assessment, highlighting its potential to transform how cities prepare for seismic events. The integration of AI into infrastructure planning marks a shift toward data-driven, predictive safety measures, setting a new benchmark for urban resilience in the face of climate and tectonic challenges.