Google details AI tools for disaster warning and response

Google is positioning its crisis resilience work around the UN report “Leveraging AI to enhance multi-hazard early warning systems,” with a focus on helping governments and international organizations prepare for natural hazards. The company has supported the UN’s Early Warnings for All initiative since its launch at COP27 and is applying AI across forecasting, real-time alerting and post-disaster response.

During the 2025 hurricane season, the U.S. National Hurricane Center used Google’s WeatherNext model to predict Hurricane Melissa’s historic Jamaican landfall five days in advance, helping Jamaica’s Met Service warn the public. In Nigeria, UN OCHA and GiveDirectly used Google river flood forecasts to trigger early action, including shelter preparation and cash transfers. Flood Hub covers 2 billion people across more than 150 countries, while pilots with WMO and hydrological agencies are testing how local streamflow data can improve regional forecasts.

Google also uses satellite imagery to track wildfire boundaries in Search and Maps across 34 countries, and three new FireSat satellites launched from Vandenberg Space Force Base in California. In 2025 alone, Google connected people with crisis information over 10 million times per day, on average. For recovery, DISHA and UNOSAT use Google models to assess building damage from satellite imagery, including preliminary damage scores for over 385,000 buildings after Hurricane Melissa in Jamaica.