15 Minutes That Decide Whether a Fire Stays Small

AI cameras are catching wildfires before the first phone call comes in. What happens after that still needs humans.

Welcome Back,

The pictures are familiar and terrifying, fires being tackled by firefighters and planes dropping water. Hundreds of thousands of people have been forced to leave their homes across France, Spain, and Portugal this summer as wildfires tore through forests and countryside, according to BBC reporting via AOL. In France alone, four firefighters have already lost their lives this summer, with another 150 injured.

So why are humans, and 30-year-old planes, still doing most of the work? If drones can operate in war zones, why can't they replace firefighters? Chief sergeant Guillaume Millet, a firefighter with 24 years of experience who now heads the Fire and Rescue Service branch of a French firefighters' union, has a clear answer, and it's more nuanced than either "AI is solving this" or "nothing has changed."

Today we look at how AI cameras are already replacing human lookout towers, why 15 minutes can be the difference between a controlled fire and a catastrophe, what the next generation of detection tech looks like, and why firefighters themselves say the job of actually putting out a fire hasn't changed since the 1670s.

📹 The Watch Towers Are Empty Now

Millet works in the Gironde, the region worst hit by wildfires in France this summer. He says the techniques for actually tackling fire haven't changed much, but detection technology has changed enormously. "The big change, especially in my département, is that now we've got cameras assisted by artificial intelligence to watch the forest and see where fires are starting," Millet says.

Those AI-assisted cameras have physically replaced the human lookouts once stationed in watch towers in every town throughout the summer months, jobs that were often filled by students. The switch has reduced false alarms significantly, since an AI-assisted camera can distinguish with almost total certainty between smoke and the dust kicked up by, say, a tractor working a field. The cameras also have serious range, detecting fires up to 20 kilometers, roughly 12 miles, away.

In the Gironde, the system was installed by a company called Midgard. In other parts of France, a separate French startup called FireTracking runs the same kind of network, and its founder gave a much more granular picture of what actually happens the moment a camera spots something.

AI-assisted cameras have replaced the human lookouts once stationed in watch towers across France.

⏱️ The 15 Minutes That Matter Most

"As soon as the system detects a fire, there's an alert on the firefighters' telephones. They open the application and see what the camera is seeing with a 40 times zoom," explains FireTracking's CEO Jean-Simon Chaudier. Next to that real-time video sits a map showing the exact location of the fire, the position of any nearby houses, and a live simulation of how it's expected to spread minute by minute, based on current weather and humidity.

Chaudier backs that up with a concrete recent example. "For the past 10 days, for example, we have the feedback from a firefighter that our AI detected seven fires 15 minutes before the first phone call. If you can help a firefighter to save 15 minutes, it's the difference between a catastrophe and a fire which can be limited to two or three hectares," he says.

💶 What It Actually Costs to Watch a Region

FireTracking's cameras currently survey a million hectares, mainly across France. In the French département of Indre-et-Loire, the company installed 11 detection sites covering the whole 6,000 square kilometer area, roughly 2,300 square miles, mainly agricultural land. Total cost: €1.2 million, about $1.4 million or £1 million.

That agricultural framing matters, crop fires alone are a huge and expensive problem in that département, with roughly 300 a year. A detection network built primarily for wildfire ends up paying dividends across a much broader set of fire risks than forests alone.

Much of the area covered by FireTracking's network is agricultural land, where crop fires are a routine, costly problem.

🚁 The Next Wave: Drones, Gas Sensors, and Solar Patrols

Camera networks are just the current generation. The French civil protection service runs a program called Condor, testing solar-powered drones that carry out continuous patrols using electro-optical cameras and AI, extending detection coverage without needing to be tethered to fixed camera towers.

That's a genuinely different detection philosophy than cameras, catching the chemical precursors of combustion rather than waiting for visible smoke. On the response side, a French startup called Daidelos is trying to raise investment for a heavy-lifting, petrol-engine drone that could hover over a fire and spray it with 1.2 tonnes of water, while other companies are developing drones that drop grenades releasing a chemical powder that smothers, cools, and interrupts a fire's combustion chemistry on contact.

The hope, according to the reporting, is that this could eventually form an autonomous first response, detection and drone activation happening so quickly that fires get extinguished right as they're starting. Notably, even the people building this technology are clear that it wouldn't do anything against a wildfire that's already established, this is specifically about the earliest possible moments.

🧑‍🚒 Why Firefighters Say the Job Itself Hasn't Changed

For all the genuine progress in detection, Millet is unambiguous about where technology's usefulness currently ends. "In fact, the usefulness of technology is very limited in this area," he says. "It can help us with detection but it's not technology that puts out fires; you need people to pull the hoses through the forest to attack them."

Even a Canadair, the amphibious plane that scoops up water from rivers, lakes, or the sea to dump on flames, has real limits. "When a Canadair drops 6,000 litres of water on a wildfire it reduces its intensity but then firefighters on the ground have to finish what the Canadair started," Millet says. Dash planes that spray orange ammonium phosphate fire retardant ahead of flames work similarly, they buy time and slow ignition, but they don't replace the brawn and water brought by ground crews.

A Canadair plane can drop 6,000 litres of water in a single pass, but ground crews still have to finish the job.

🧠 AI Spotlight Analysis

What's striking about this story is how cleanly it separates two different problems that easily get conflated, detecting a fire and extinguishing one. AI has made genuinely dramatic progress on the first, real-time cameras with 20-kilometer range, minute-by-minute spread simulations, and a documented 15-minute head start over a human phone call. That's not incremental, it's a fundamentally different early-warning system than towers staffed by students with binoculars.

The second problem, actually stopping a fire once it exists, has barely budged. Planes are decades old, hoses still need human hands, and the drones being developed for direct suppression haven't reached the field yet even in testing. That gap between detection and suppression is exactly where the real opportunity for future investment sits.

That's a genuinely humbling frame for a technology story, nearly 350 years of firefighting technique, and AI's role so far is entirely upstream of the moment the fire actually gets fought. It's a real, life-saving contribution. It's also, by the estimate of the people doing the work, still a limited one.

🔗 Sources and Further Reading

Thanks for reading AI Spotlight.

Our mission is simple: deliver clear, trustworthy, and actionable AI insights that help professionals stay ahead without the hype.