Book a demo

The privacy posture, by architecture

These are not policy paragraphs that could be quietly updated later. They are how the pipeline is built.

Face matching is not wired in this product

Face matching is a capability we are building and is not turned on today — the recognizer is not yet wired: it holds no face-recognition model weights, and no face template is computed from a photo. Photo finding uses a permission-checked roster lookup instead. A parent reviewing the opt-in sees an opaque reference, never the template itself. Withdrawing the opt-in stops the matching.

One thing we will not overstate

The step that destroys the stored template is not finished, and we are not going to tell a parent it runs nightly when it does not. The cleanup job is built to refuse: when it cannot actually destroy a template, it halts and raises an alert rather than mark it deleted. We would rather leave that alert standing than record a deletion we cannot show you. When it can be demonstrated end to end, this page will say so plainly.

No outside AI or photo company

Photos and any face data run on our own private system. A portrait is never routed to a third-party AI service, an ad network, or a general-purpose outside photo lab’s storage. Editing and storage stay with us.

Per-center tenant isolation

A center’s roster, photos, and consent records are never visible to another center’s session. This is enforced at the data layer, forming a single-center FERPA-style privacy wall, not a convention a route could forget to apply.

Scoped staff roles

Center staff see only what their role permits — a front-desk view of the roster is not the same as a full consent-and-order view. Access is scoped, not all-or-nothing.