Flox raises €1.4M for AI that spots wildlife on airport runways
What's the deal? Flox IntelligenceDealroom has a profile for this one. Try Dealroom →, a Madrid-based startup founded in 2024, has raised €1.4 million in a seed round led by Kibo Ventures. The money will fund its AI system for detecting and managing wildlife at airports.
What does it do? Flox uses computer vision and animal behaviour models to flag birds and mammals near runways before they collide with aircraft. The system now processes images from more than 1,200 cameras at once, with 800 milliseconds between capture and alert.
What's different? Conventional radar detects the mass and speed of flying objects but cannot classify species or predict short-term trajectory. Flox combines thermal cameras, optical cameras, and local weather data to judge whether a bird is landing, flocking, or settling on the tarmac — the detail that determines actual risk.
Its model was trained on 4.7 million annotated images across 87 bird species, plus a "mammals in the manoeuvring area" category for foxes, rabbits, and boars. Species-classification accuracy reaches 91.3% for the 20 highest-risk species at European airports.
Why now? The problem is costly and growing. Aircraft–wildlife collisions caused 263 declared incidents at European airports in 2025, with repair and delay costs the sector estimates at more than €800 million a year worldwide.
Flox has spent twelve months running a pilot at Madrid's Barajas airport and two regional airports. Current airport systems — human patrols, acoustic deterrents, and bird-tracking radar — offer only partial coverage and react rather than predict.
How does it make money? Flox runs a hybrid model. The detection software is sold as SaaS priced per active camera per month and can plug into an airport's existing cameras; for older infrastructure, the company rents out its own thermal cameras with the cost folded into the subscription.
The signal: Flox is a concrete case of the AI agent boom moving into specialised, measurable applications — not general-purpose chatbots, but models trained on domain-specific data. The round sits in roughly the 68th percentile by size, modest in absolute terms but notable for a vertical tackling a problem with a clear price tag.
Read more: wwwhatsnew.com
Image credit: Flox Intelligence
