Privacy
Your recordings, and what happens to them
Written for this review build. It describes what the software actually does today, not what a finished product would do.
This review build does not accept real candidate data
Please do not record confidential project information or anyone's personal details here. No personal data has been imported into this build and all sample records are synthetic.
Where your audio is stored
Recordings you make stay in your own browser, in its local database, on the device you used. They are not uploaded to a Pr-ed server in this build, they are not shared with a mentor account belonging to anyone else, and they are not used to train any model. If you clear your browser storage or use the reset control in settings, they are gone permanently — there is no copy elsewhere to restore from.
Transcription
If your browser offers speech recognition and you explicitly switch it on, your browser may send the audio to a service operated by the browser vendor to convert it to text. That happens outside Pr-ed's control and needs an internet connection. If it is unavailable, refused or fails, we keep your actual audio and mark the answer “needs transcription” — we never substitute invented text for what you said.
Deleting things
You can delete an individual recording, delete a whole session, or wipe every trace of your local data from settings. Deleting a recording removes the stored audio, not only the reference to it.
What production would require
Before real candidates use this, a production deployment would need authenticated tenant-scoped storage, row-level access control, private object storage with short-lived links, explicit consent capture, defined retention and deletion periods, an audit trail, and approval of any cross-border processing. Those are design requirements set out in the rebuild brief, not things this review build has implemented. Processing would be handled in line with POPIA obligations; that is a target for the production system, not a compliance claim about this one.