A system developed by INSAIT uses artificial intelligence to predict wildfire risk across Europe, providing governments with data for prevention and human-understandable explanations. FireScope AI is the first system of its kind and has been developed by the Institute for Computer Science, Artificial Intelligence and Technology (INSAIT) at Sofia University “St. Kliment Ohridski”.
INSAIT’s development comes at a time of increasing wildfire risk in Europe. According to the European Environment Agency, wildfires cause approximately 2.5 billion euro in damages annually across the EU, while the World Meteorological Organization projects that wildfire risk in Europe will continue to increase due to climate change. Against this backdrop, INSAIT’s breakthrough in such a societally critical area demonstrates how AI research can have a direct impact on protecting people, infrastructure, and nature - not only in Europe, but globally.
INSAIT is also launching the first public high-resolution AI map of long-term wildfire risk across Europe. The map enables institutions and governments to more effectively target prevention, preparedness, land-use planning, and climate adaptation measures toward areas with the highest risk of fire occurrence and severe damage.
INSAIT’s system outperforms both general-purpose AI models such as GPT-5 and conventional methods such as the Fire Weather Index.
The system works by combining satellite imagery and climate data to generate continuous wildfire risk maps. To produce robust, expert-level fire risk estimates, it uses these inputs to infer vegetation density and dryness, seasonal drought trends, wind patterns, and complex interactions between heatwaves, terrain, and vegetation.
FireScope AI uses a novel “reasoning-to-generation” pipeline: a thinking model we trained to explicitly reason over Sentinel-2 imagery and climate variables and a vision model that turns its signal into dense wildfire risk rasters.
The project is built in collaboration with researchers from ETH Zurich and has been accepted for presentation at CVPR 2026 - the world’s most prestigious conference in computer vision and among the most important conferences in artificial intelligence.
The project was developed by Mario Markov, Stefan Maria Ailuro, Luc Van Gool, Konrad Schindler, and Danda Paudel.
The interactive map for Europe is available at: www.firescope.ai .