“Are your evacuation drills a herd of people all headed to the front door because that’s closest to the muster point, or do they reflect the chaos that will accompany a real event?” – Ken Hayes, Polaris MEP Project Manager
The National Safety Month spotlight on prevention makes this a timely moment to share how new technologies could improve emergency outcomes in buildings and industrial facilities. The National Institute of Standards and Technology (NIST) has developed an artificial intelligence model, called Safe Step, that is designed to help people evacuate more safely during fires by continuously steering them toward routes with the lowest danger, not just the closest exit. The model uses reinforcement learning and data from fire simulations to anticipate how flames and toxic smoke will spread, then evaluates each possible route using a fire safety metric known as the fractional effective dose (FED) of hazardous gases. In a smart building equipped with sensors and dynamic electronic exit signs, Safe Step could update guidance in real time—redirecting occupants away from a hallway about to fill with smoke and toward a more distant but safer exit.
NIST researchers tested Safe Step in simulated single-story building layouts and found that it consistently generated safer evacuation paths than conventional shortest-route algorithms, especially in more complex floor plans where conditions change quickly. The team is now working to extend the approach to multi-level buildings and to model the behavior of many evacuees at once, with the long-term goal of integrating this technology into real-world emergency systems over the next five to ten years, pending rigorous testing and regulatory approval. For safety leaders and facility managers looking to elevate their programs during National Safety Month and beyond, Safe Step points to a future where AI-enhanced infrastructure supports faster, more informed evacuation decisions and potentially saves lives during fire emergencies.