AI Fall Detection Keeps Grandma Safe, if She’s OK With Being Watched

In the Dutch municipality of Waalre, 10 older adults are now living under the quiet watch of artificial intelligence. Ceiling-mounted sensors from Kepler Vision Technologies scan their homes continuously, feeding an AI trained to distinguish a fall from a sit-down and automatically push a notification to family members or emergency contacts when the algorithm flags an incident. Depending on how you feel about surveillance tech, that either sounds like a great way to protect independent older people who live alone or like a dystopian nightmare. The pitch, at least on paper and given the alternative, leans toward the former.

According to Statistics Netherlands, just over a quarter of the Dutch population will be over 65 by 2040, yet the country’s care infrastructure is not growing at nearly the same rate. This isn’t a problem unique to the Netherlands. In the US, we’ll reach similar numbers by 2050. Japan’s over 60 population is already around 30% today and the World Health Organization predicts that the global population over 60 is expected to nearly double by 2050. That means there’s more pressure for older adults to manage independently at home, for longer, with less institutional support every year. Falling — more specifically, lying undiscovered after a fall — is one of the more dangerous consequences of this unfortunate calculus, but the faster someone is found after a fall, the better their chances of recovery are.

This Dutch pilot, run through a collaboration between connectivity provider WeConnect, care network Leefsamen, and Brainport region partners, is designed for people already at elevated fall risk who want to stay in their own homes. The hardware and software are similar to the AI fall-detection systems Kepler has been running in nursing facilities for some time. So, this first application in private residences is a logical extension, not necessarily a conceptual leap.

And yet, the idea of an all-seeing eye inside a home seems, well, weird.

A sensor that can reliably detect the movement pattern of a fall can, by definition, detect a great deal else about how someone moves through their home — when they get up at night, how often they visit the bathroom, whether their gait is changing. Even if the system is designed to suppress that data, the infrastructure for collecting it exists. If the pilot scales, what happens when the commercial incentives of the companies involved diverge from the privacy interests of a 78-year-old who signed a consent form she may not have fully understood? What happens in the event of a data breach?

These aren’t hypothetical concerns — heck, they aren’t even limited to this pilot program, since the tech is already monitoring more than “15,000 elderly people around the clock” in care facilities, according to Kepler’s release. The partner companies have made the familiar pledges to protect privacy with Kepler specifying compliance with international information security standards which is a little reassuring, but data breaches happen.

None of this makes the technology bad; it’s just complicated. For someone who’s living alone, the choice may not be between AI monitoring and unmonitored freedom; it may be a choice between AI monitoring and a fall that goes undiscovered for two days. Framed that way, the sensor in the hallway starts to look less like surveillance and more like a smoke detector with better software.