#1034 2025 · Sapeurs-Pompiers de l'Hérault (SDIS 34), Havas Play, Twitch · Public-safety / wildfire detection (livestreaming medium)
Turned Twitch's wait screens into a wildfire watch whose viewers dispatch firefighters
the problem
France had hundreds of forest cameras but too few operators to watch them, so fires won when nobody spotted first smoke
background
Wildfire management in France depends on catching a fire in the first minutes after ignition, when it can still be stopped, but the detection network has a staffing bottleneck: hundreds of surveillance cameras watch fire-prone forests (in the pilot, Hérault's PARATRONIC 'doubt-clearing' cameras via the department's CODIS emergency operations centre), and there are nowhere near enough human operators to stare at every live feed continuously. Meanwhile millions of people already sit in front of connected screens every day paying focused attention to nothing in particular — most visibly on Twitch, where viewers wait on a 'Starting Soon' screen before a stream begins.
The obvious response — hire more operators, or build better computer-vision detection — either spends scarce emergency budget staffing the problem or waits for an algorithm that is still error-prone. Neither uses the one resource that is already free and enormous: distributed human attention.
what everyone would do
Every default answer spends scarce budget on the bottleneck: hire more operators and the emergency service runs an expensive 24/7 watch; buy or build computer-vision detection and you wait on imperfect algorithms and miss the edge cases humans catch; or push out a 'be vigilant' appeal and get a trickle of sporadic, unverifiable tips. None of these captures the one abundant, free resource — the millions who already stare at screens — because none re-prices their existing idle attention as work.
what they saw
Viewers on a 'Starting Soon' screen already watch a screen for free; only the object is useless. Swap the dead frame for a live feed and that wasted attention becomes detection, crowd consensus replacing an operator.
the move
Havas Play replaced four major French streamers' idle 'Starting Soon' waiting screens with live feeds from forest-surveillance cameras, teaching viewers to spot early smoke. Hundreds of thousands of viewers watched as they normally would; when several reported the same suspicious smoke in chat, their simultaneous flags automatically alerted the CODIS fire operations centre, which took over verification and response. The concept then spread voluntarily to other creators, adding 2,190 hours of live surveillance to the pilot season.
why it works
It works because the cost of being a spotter drops to zero and hides inside a habit: viewers watch what they were going to watch anyway, so there is no volunteering, no shift to commit to, no training beyond a line about what smoke looks like. Because no individual is asked to act or be accountable, the crowd participates without friction; the fire service is protected from false alarms by requiring several independent reports before auto-alerting CODIS, and the emergency operation centre keeps the real decision. Scaling is organic because each participating creator brings their own existing audience, so surveillance capacity grows without a single hire, and it broadens to whoever joins — the concept spread from four streamers to many voluntarily.
the payoff
Campaign self-reports 600,000+ followers, 2,190 surveillance hours, 5 fires, ~1,700 hectares protected in Hérault pilot; figures unverified.
where it breaks
It fails where there is no idle, already-captive audience that can see the hazard — a remote site with no connected screen, a danger not visible on camera, or a public that does not gather in front of one shared medium. It needs the reporting threshold tuned: too low and false alarms flood the emergency centre; too high and genuine first-smoke alerts get missed. And the crowdsourced verdict is only as good as the feed quality and the instructions, so it degrades in niches where the observers are too few, too distracted, or unable to judge what they see.
what came after
Won a Cannes Lions 2026 Lion, and the fire service has framed it as scaling nationally and to other European countries, positioning dormant screen-waiting time as a reusable detection capacity rather than a one-off stunt.
references
- [1]France 3 Régions — Une tour numérique de surveillance citoyenne : FireCatchersFrance 3 (France Télévisions / public broadcaster), 2025france3-regions.franceinfo.fr
- [2]Gens d'Internet — Les Sapeurs Pompiers lancent un dispositif de vigilance pour de potentiels incendies sur TwitchGens d'Internet, 2025gensdinternet.fr
- [3]Paratronic — FireCatchers : Twitch et PARATRONIC contre les incendiesParatronic (surveillance-camera supplier), 2025paratronic.com
Widely retold, only partly documented. Filed as hearsay.