Google is expanding its deployment of artificial intelligence to strengthen global crisis resilience, partnering with governments, humanitarian agencies, and scientific institutions to enhance disaster forecasting, public alerting, and post-disaster response. Across the 2025 to 2026 operational cycle, these initiatives have demonstrated measurable improvements in early warning capabilities and recovery efficiency worldwide. In forecasting and preparedness, Google’s WeatherNext model provided the U.S. National Hurricane Center with a five-day advance prediction of Hurricane Melissa’s landfall in Jamaica, allowing local authorities to issue timely public warnings. In Nigeria, UN OCHA’s Floods Anticipatory Action Programme and NGO GiveDirectly utilized Google’s river flood forecasts to trigger early interventions, including shelter preparation and pre-emptive cash transfers that enabled families to evacuate and secure property ahead of severe flooding. These forecasting tools are accessible through Flood Hub, covering 2 billion individuals across more than 150 high-risk countries. To refine model accuracy, Google collaborated with the World Meteorological Organization and hydrological agencies in Czechia, Nigeria, Uruguay, and Vietnam to integrate local streamflow data into global AI frameworks, significantly improving predictions for ungauged watersheds. Supporting broader research, Google open-sourced its Groundsource urban flash flood dataset and hydrology modeling framework, which the Czech Hydrometeorological Institute has successfully adapted for national workflows. For wildfire management, Google tracks boundaries across 34 countries using satellite imagery and recently deployed three new FireSat satellites alongside Earth Fire Alliance and Muon Space to accelerate early detection globally. During active crises, Google’s Public Alerts system, powered by Common Alerting Protocol feeds from authorities in over 90 nations, distributes verified warnings across Search, Maps, and Android devices. The platform averaged over 10 million daily crisis interactions throughout 2025. Notable implementations include Android’s earthquake alert network, which converted millions of devices into seismometers to provide advance shaking warnings to users outside the epicenter during a recent Venezuelan earthquake. Post-disaster response has similarly been optimized through AI-driven damage assessment. In partnership with the UN Satellite Centre and Data Insights for Social and Humanitarian Action, Google integrated Open Buildings and building damage models to rapidly analyze satellite imagery. This workflow has been activated 11 times since deployment, dramatically reducing manual assessment timelines. Following Hurricane Melissa, the system generated preliminary damage scores for more than 385,000 Jamaican buildings, directly informing recovery operations. Similarly, after February 2026 flooding in Colombia, AI-derived building maps were cross-referenced with radar imagery to rapidly map infrastructure damage, enabling streamlined humanitarian planning for UN agencies and national authorities. By unifying predictive modeling, real-time alert infrastructure, and automated damage assessment, Google’s AI crisis initiatives are establishing a scalable blueprint for disaster management, transforming how governments and aid organizations anticipate, respond to, and recover from natural hazards.