“Refusing means your data will be deleted.” That kind of phrasing in a health app causes panic — even when it doesn’t mean what it sounds like.

What the 7.00 Update Asked Users

With Samsung Health v7.00.5.009, users encountered a consent dialog requesting permission to share four data categories: general health and wellness, medication, health records, and menstrual cycle data. The dialog indicated that refusing would result in those data being deleted.

The problem: the dialog didn’t specify which data, or what the sharing was for. For a user who has tracked their health for years in the app, reading “your data will be deleted” is alarming enough to provoke an immediate reaction — and that’s exactly what happened across forums and social media.

What Samsung Clarified After the Confusion

Samsung confirmed after the fact that the dialog concerned only data used for AI model training — not personal tracking data stored locally or in the Samsung account. In other words: refusing AI-training data sharing does not delete your personal health history. Your heart rate, steps, sleep, cycle data — all of it remains intact.

What Samsung deletes when consent is refused is the copy it would have used to improve its health AI algorithms. A real distinction, but one the original dialog made no effort to communicate.

The Trust Problem This Reveals

This confusion isn’t an isolated communication accident. It illustrates a structural problem in how tech companies communicate about data practices: informed consent is routinely sacrificed in favor of generic language that maximizes acceptances or minimizes questions.

Samsung isn’t exceptional here — Meta, Google, and Apple have all had similar episodes. But in health specifically, the bar should be higher: menstrual cycle data, health status, and medication records are among the most sensitive data a user shares with any app.

The open question: how many users refused consent without realizing they were losing nothing from their personal tracking? And how many accepted without understanding their health data was going toward model training?