Health data is the most sensitive category most people will ever hand to software. These are the questions worth asking, and what a good answer sounds like.
1. Is my data used to train your model?
A good answer distinguishes between learning how to read a document and learning from your document. Ours: we do not train on information that could identify you. What we improve is how a value and a printed range are read, not who you are.
2. Who inside the company can see my results?
“Nobody” is rarely true and usually means the question was not understood. A real answer describes access control and why some access exists.
3. Can I delete everything, and does deletion mean deletion?
Look for a route that does not require emailing support, and for honesty about backups and the window before they expire.
4. Is my data sold, shared or made available for eligibility decisions?
This is the one that matters most and gets the vaguest answers. Health results reaching an insurer or an employer is a different order of harm from an advertising profile. Ours: never, and it is not a business line we are keeping open.
5. What happens when a family member uses my account?
Most health products are built for one person and then have family bolted on. Ask whether profiles are genuinely separated or whether everyone shares one record — the second is common and is how a teenager’s results end up visible to a parent, or the reverse.
The meta-question
Ask what the business model is. If the answer is not clearly “people pay us”, your data is a candidate for being the product, whatever the policy says.
This article explains how our technology works. It is health information, not medical advice.