Face apps are fun, and most people tap "upload" without a second thought. That's fine for a photo of your lunch. A face is different: you can change a password, but you can't change your face. Here's a short, practical checklist for any app or site that asks for your face, followed by an honest account of how FaceTest handles it.
Why face data is a special case
Most privacy laws treat biometric data, including facial geometry, as a sensitive category. The EU's GDPR lists biometric data used for identification among its special categories; Korea's Personal Information Protection Act treats biometric information as sensitive information requiring separate consent; several US states have dedicated biometric statutes. The legal detail varies, but the reasoning is the same everywhere: a face template can identify you across contexts, it's permanent, and a leak can't be fixed by resetting anything.
The checklist
- Does the photo leave the device? This is the biggest question. If analysis happens on a server, your photo is transmitted and, at least briefly, stored. If it happens in the browser or app, it may never leave.
- What is stored, and for how long? "We delete photos after processing" is good. "We may retain images to improve our service" means your face is training data.
- Is a face template stored? Even without the photo, a stored embedding is biometric data. Look for words like "facial feature vector", "template", or "faceprint".
- Is an account required? An account links your face to an identity. A tool that works without one has less to leak.
- Who are the third parties? Analytics and ad networks are common; a policy that shares "user content" with partners is not.
- Can you delete? If something is stored, there should be a way to remove it.
- Is sharing opt-in? A result that's public by default, or a share card that includes your photo, is a choice you should get to make.
What "on-device" actually means, and how to verify it
An on-device (or in-browser) app downloads the model to your phone or computer and runs it there. Your photo is read from disk into memory, processed, and discarded when the page closes. You don't have to take that on faith. On a desktop browser, open developer tools, switch to the Network tab, and run the test. Any upload of your image would show up as a request carrying an image or a large payload. If you see the model files download and nothing go back up, the claim holds.
How FaceTest handles your photo
- The neural network runs in your browser through TensorFlow.js. Your photo is never uploaded, stored, or transmitted. You can verify this with the method above.
- The comparison targets (character vectors) download to you; your vector never goes the other way.
- No account exists, so there is nothing to link a face to.
- When you share a result, you create a short link that points to the character you matched. The share card shows the character's image and the score, not your photo.
- We count how many times a test is played and shared, and which result was reached, as plain numbers. That feeds the result-distribution stats on each test. It contains no image and no personal identifier.
- Analytics on the site uses standard page-view tools; it never sees your image because your image never leaves the page.
Good habits for any face app
Revoke camera permissions you no longer use. Prefer tools that work without sign-up. Skip apps whose policy mentions training on your images unless you're comfortable with that. And when a policy is vague about biometrics, treat the vagueness itself as the answer. See how the matching works for why a face test doesn't need your photo to leave your device in the first place.