As someone who’s witnessed dementia up close, I’m aware how devastating the illness can be.  It strips identity and ability away from its victims, and it is extremely challenging to manage for caregivers.  They can’t just park people in front of televisions and making sure they take their meds.  They also have to make sure they don’t fall, protect them from scams and frauds, deal with personality changes, monitor that they don’t wander off, and manage a whole range of other issues.  It’s not unlike supervising a toddler.

It’s even more challenging in our present society.  Decades ago, children probably lived in the same city as their aging parents.  Today, they’re more likely to be spread out geographically, so when mom or dad begins to show signs of impairment, those who can help are often on the other side of the country.  This means relocation – itself a disorienting challenge for dementia patients – or putting loved ones in an assisted living/memory care home.

And those can easily cost $10,000 or more per month.

AI to the Rescue

Wouldn’t it be great if you could drop something into your loved one’s home and let AI monitor them, ensure they take their meds, and summon help if they fall?

Recently while reading one of the dementia care subs on Reddit, I’ve noticed comments from someone at Guardian Angel AI promoting their product.  Once they launch, they will offer AI-powered cameras that you can place in your loved one’s home.  These systems monitor your loved one (LO), learn their habits and needs, and can alert if there are issues such as falls or wandering.  They can also power medication dispensers, analyze patterns, and record a history for clinician analysis.

This is one of the worst ideas I’ve ever encountered.

Tech Bros Are Not Caregivers

Although the site claims it was “Designed alongside geriatricians and dementia specialists” there is no information on their site who these medical professionals people are.  There is a “Caregiver Field Guide” PDF that is available for download, but it’s obviously AI slop and is 20 pages of description of dementia symptoms, not how to implement their technology.

Indeed, the entire site feels like it was prompted.  This does not feel like a well-funded health care startup where someone in the medical field had an idea based on their clinical experience and is bringing it to market.  Rather, it’s apparently a technical guy who had an idea of how to make an AI product.  That’s the wrong side to approach a solution from.

Let’s talk about some important issues.  Keep in mind that at no point does Guardian Angel say they have a call center of people that monitor alerts.  Everything is AI-powered.

Legal Liability

The liability profile here looks unusually serious because Guardian Angel is positioning itself in the gap between a consumer software product, an elder-care monitoring service, and a safety/emergency system. The homepage says the system “quietly learns your parent’s day” and alerts family when “something genuinely needs you.” That sounds reassuring commercially, but legally it creates an expectation that the product will recognize dangerous deviations.

Imagine the product learns that an 82-year-old normally gets up at 7:30. One morning she falls at 6:45 and remains on the floor for eleven hours.  The family alleges: “We bought GuardianAngel specifically so this wouldn’t happen.”  That’s a potentially brutal negligence/product-liability fact pattern.

In court, a Terms of Service disclaimer saying “this isn’t an emergency service” does not magically erase the duty created by the actual product design, representations, onboarding, and foreseeable reliance on the system.  The closer their marketing gets to “we notice when your loved one needs help” – and that is the main marketing pitch – the worse the reliance problem becomes.

Regulation Risk

The customer here may be the child who wants to monitor mom and dad, but the data subject is not the customer – it’s the loved one.  Can they really consent to being monitored if they have dementia?  Leaving that aside, what about home-health aides, nurses, cleaners, grandchildren, neighbors, delivery workers and guests?  They never consented.

Now granted, lots of people have Ring/Blink/whatever cameras in their home.  But there’s a twist here: we’re talking about healthcare data.

Because of all the AI processing of the video and audio, the data is inevitably passed through other hands.  So if a nurse walks into the monitored LO’s home, everything he or she does is recorded and probably send to some cloud service for processing and analysis.  During the visit, all kinds of info is captured, including a ton of biometrics.  Some states have laws that go way beyond HIPAA, and they’re all about how healthcare data is handled.  The State of Washington’s law is called “My Health, My Data” if that gives you any indication.  Inferencing health patterns is specifically what these laws are designed to regulate.  Guardian Angel’s statement on HIPAA compliance is laughably short and doesn’t even mention such laws.

Security Risk

Needless to say, if Guardian Angel’s systems are ever compromised, it’s a goldmine for attorneys (assuming that GA has any assets).  You’ve got video data, healthcare data, and the usual customer data all in one golden package.  And as a bonus, we’re talking about seniors who are often very vulnerable to scams.

An attacker might learn: when an elderly person sleeps, when they’re alone, when caregivers arrive, when family is present, whether they’re mobile, what’s in the house, etc.

It’s a Medical Device, Guys

Their site says

GuardianAngel AI™ is not a medical device and should not be used as a substitute for professional medical advice, diagnosis, or treatment.

If the product detects health deterioration or events indicating someone requires intervention, that sure sounds like a medical device.  At a minimum, it sure needs to be legally vetted because it’s right on the edge.

And even if it’s not, the negligence case for the plaintiff is going to ask if the safety-analysis algorithm was tested.  And by tested I mean sensitivity/specificity by event category, false-negative rates, false-positive rates, confidence thresholds, demographic performance, testing on mobility aids, testing across lighting/home configurations, sensor-failure detection, network-failure behavior, model-version validation and post-market incident monitoring.  Does the GA developer really have all that data?

False Dispatches

I couldn’t find on the site if it interacts with the 911/EMS system.

If so, that’s another huge liability.  If you’re spamming the customer at 2am, that’s a hassle for the customer.  If you’re spamming the 911 operator at 2am, that’s a completely different thing.

Actually, it’s thousands of different things, because you have to comply with every jurisdiction’s rules.

In Short

GuardianAngel shouldn’t think of itself legally as “an AI SaaS company for families.”

It should assume regulators, insurers and eventually plaintiffs will characterize it as a safety-critical elder-monitoring company making automated judgments from sensitive in-home behavioral data.

And that is a whole other level of ethical and legal responsibility and exposure.  The fundamental issue is that one false negative can produce a seven-figure wrongful-death claim, while one systemic privacy defect can produce a regulator or class-action problem across every installed household simultaneously. That’s an unusually dangerous combination of severity risk and scale risk.

I won’t be investing.

raindog308

raindog308 is a longtime community LETizen, technical writer, and self-described techno polymath. With deep roots in the *nix world, he has a passion for systems both modern and vintage, ranging from Unix, Perl, Python, and Golang to shell scripting and mainframe-era operating systems like MVS. He’s equally comfortable with relational database systems, having spent years working with Oracle, PostgreSQL, and MySQL.

As an avid user of LowEndBox providers, raindog308 runs an empire of LEBs, from tiny boxes for VPNs, to mid-sized instances for application hosting, and heavyweight servers for data storage and complex databases. He brings both technical rigor and real-world experience to every piece he writes.

Beyond the command line, raindog308 has a life-long love of German Shepherd Dogs, high-quality knives, target shooting, theology, tabletop RPGs, playing guitar, and hiking in deep, quiet forests.

His goal with every article is to help users, from beginners to seasoned sysadmins, get more value, performance, and enjoyment out of their infrastructure.

You can find him daily in the forums at LowEndTalk under the handle @raindog308.