In the past, communities relied on traditional methods such as 911 calls to identify wildfires. Once alerted, first responders rushed into fast-moving wildfires with almost no real-time data or a shared map of where the fire was moving.
AI is changing that. The technology can now be used to detect fires shortly after they ignite, giving firefighters a chance to extinguish them before they rage out of control.
Here are a few ways AI is being used to prevent fires.
Pano AI uses wildfire detection stations and AI
California-based Pano AI has trained the technology to detect smoke, pinpoint its location, and provide real-time images and maps to first responders. If that sounds easy, think again. “Training AI to reliably detect wildfire smoke is an extremely difficult problem to solve,” said Sonia Kastner, CEO of Pano AI. “Smoke isn’t a fixed object. It shifts in shape, color, and opacity, depending on weather, lightning, and terrain.”
The company has positioned wildfire detection stations, equipped with cameras, in high-risk areas across 30 million acres in the U.S., Australia, and Canada. The technology uses billions of images to give AI a baseline for what “normal” looks like, so it can detect anomalies like smoke.
Pano AI has already identified hundreds of wildfires and provided firefighters with the maps and information they need to save lives. It’s done this work, in part, through an investment from Salesforce. The startup is part of the Salesforce Ventures Impact Fund, which invests in mission-driven, for-profit companies.
FireSat employs satellites and AI
Meanwhile, Google Research is part of the nonprofit Earth Fire Alliance — other partners include the Gordon & Betty Moore Foundation and Muon Space — that has developed FireSat, a satellite constellation solely dedicated to detecting and tracking wildfires.
FireSat collects advanced infrared images of Earth and will use AI to compare current images with thousands of prior images taken in the same spot. It will also factor in other data, like local weather, to determine whether there’s a fire.
The system’s first satellite was launched from California in March 2025, and quickly demonstrated its capabilities by detecting a small fire not seen by other satellite systems. When it reaches full operating capability, FireSat will provide global, high-resolution imagery that’s updated every 20 minutes, letting users detect wildfires as small as 5×5 meters, or roughly the size of a one-car garage. “Earth Fire Alliance, Google and our partners are on track to have three more satellites up and running by early 2027,” said Google.org’s Giving Lead Brian Juhyuk Lee.
FireSat will also provide scientists and wildfire authorities with an unprecedented, worldwide dataset to predict wildfires and improve wildfire behavior modeling.
