Flox Intelligence wants to replace a context-blind alarm with a device that first identifies an animal and then chooses how to respond. Roughly 150 edge units already operate across North America, Europe and Australia. The next challenge is not another recognition demonstration: it is showing that adaptive deterrence remains effective as equipment becomes transport and utility infrastructure.
The company is opening a $10 million Series A to expand toward several thousand devices, according to the funding report. That is a financing objective, not a closed round. A larger installed base will test maintenance, species coverage and support as much as the algorithm.
A local decision loop, not simply a louder siren
Flox spun out of Stockholm's KTH Royal Institute of Technology in 2020. Founder and CEO Sara Nozkova previously spent 13 years in transportation and autonomous vehicles. Her premise is that repeated noise does not solve the same problem as understanding which animal is present and how it is behaving.
Flox Edge wakes on a motion sensor, captures a camera image and runs an on-device identification model. It selects a species- and context-oriented audio response, combining computer vision with bioacoustics and behavioral observations. Local processing reduces network dependence and response delay beside remote tracks, roads or airfields.
The company varies signals to address habituation, when animals stop reacting to repeated stimuli. A fox, goose, moose or bear may respond differently, and behavior can change with feeding, migration or offspring. Correct classification alone still cannot guarantee movement; an animal may be stressed, distracted or accustomed to similar sounds.
Units attach to poles, trees or existing structures. A dashboard records deployments and events, with company-described compliance and biodiversity reporting. Customers buy hardware, then pay an annual software and AI subscription after the first year. OEM licensing offers another route; Alstom is testing integration into railway equipment.
Flox has raised about $5 million in earlier rounds and expects approximately $1.5 million revenue for 2026, Nozkova told TechCrunch. Those figures and the 150-unit footprint provide a starting business, not evidence that a several-thousand-device service network already exists.
The collision problem has different denominators by species
The FAA wildlife-strike data records about 344,000 reported strikes involving US civil aircraft from 1990 through 2025, including around 24,500 at 870 airports in 2025. The estimated annual average monetary loss is approximately $272 million over that period, not a complete account of ecological or operational harm.
Birds represent 96% of reported US civil-aircraft strikes. Waterfowl account for just 4% of strikes but 23% of damaging ones; about 70% of events occur between ground level and 500 feet. The relevant question is therefore not merely whether a device detects birds, but which species and situations it handles.
A United Boeing 737 MAX 9 struck a coyote during takeoff at Chicago O'Hare in January 2025 and returned safely, requiring inspection and disrupting passengers, as AP reported. Ground animals matter too. Alstom puts Sweden's annual animal collisions at about 5,000; SNCF says wandering animals affect more than 8,000 trains in France each year.
In Brittany, 120 animal collisions in 2023 disrupted 660 services and caused more than 15,000 delay minutes. Yet continuous fencing is costly and fragments habitat, while patrols cannot cover every corridor. Dogs, falcons, lights, lasers, pyrotechnics and habitat management remain useful, with effectiveness dependent on species, terrain and follow-up.
Airport and rail trials support narrower claims than universal safety
From 2022 to 2024 Flox worked with Swedavia at Malmö, Kiruna and Umeå airports. The company's trial account reports 100% deterrence across more than 200 animals, including predatory birds, geese, foxes and ungulates. This is a vendor-reported result under specific conditions, not an independently established rate for every airport.
At Mora Airport, Flox says magpies, crows and other animals continued avoiding protected runway areas after six months, with less reliance on patrols and improvised scaring. That provides a longer example, but buyers still need success definitions, missed detections, return times, species breakdowns and season-by-season performance.
Flox and Alstom first put AI cameras on trains to observe wildlife on Swedish lines; a second phase added sound deterrence in April 2026. The rail-trial report lists moose, roe deer, foxes, wild boar, farm animals and birds.
Alstom said recognition was particularly good for farm animals, crows and pigeons, while moose and roe deer needed more training. That distinction matters because dangerous species can also be harder to identify. Detection counts establish presence, not reduced collisions, and moving trains present different timing and geometry from stationary units.
Flox advertises response below two seconds, but the safety margin includes camera placement, distance and the animal's direction—not inference time alone. Its statements about learning also need operational clarity: periodic model updates, redesigned sound libraries and autonomous local adaptation imply different oversight requirements.
Adaptive audio must fit a layered safety and welfare plan
The FAA favors integrated wildlife management, including habitats, food and water sources, vegetation, fencing, dispersal and removal where needed. Flox can be another layer, not a replacement for inspection, reporting or emergency response. Comparable automated approaches include Robin Radar's Scarecrow species-specific distress calls and iCHASE's vision-directed moving laser.
False negatives leave hazards in place; false positives produce unnecessary sound, distracting alerts and lost trust. Rain, snow, darkness, glare and vegetation challenge cameras. A model that works around a Swedish airport may not transfer unchanged to mammals beside an Australian road.
Nonlethal does not automatically mean harmless. Sound intensity, exposure, effects on non-target species and nearby people, and where animals move next all matter. Driving an animal from one protected area into another hazard is not a successful safety outcome. Independent long-term observation is stronger than a selected success story against habituation.
Production adds dirty lenses, weak batteries, degraded speakers and changing vegetation. Operators need authority over model and sound-library changes, signed and controlled updates, and event records. Local processing reduces connectivity dependence but does not remove the risks in dashboards, account access or fleet software distribution.
The infrastructure case is renewal plus fewer incidents
Three developments would materially strengthen the case: independent species-level detection and deterrence results across difficult seasons; Alstom outcomes showing fewer collisions or high-risk incursions; and repeat orders, renewals and wider OEM adoption as deployments expand. A fundraising target or a busy dashboard cannot substitute for those results.
Flox also wants its perception layer in cars, drones, tractors and outdoor robots. Warning a driver, changing a route or controlling a vehicle are different safety responsibilities. A farm deterrent, runway unit and vehicle-control function cannot share one undifferentiated acceptable-error standard.
The plausible opportunity is a system that helps operators decide where, when and how to intervene, while its event history improves fencing, inspections and habitat planning. To establish that role, Flox must show sustained reductions in incidents, delays, patrol work and animal harm—not simply more accurate labels or a larger number of detections.