Biometric Data Retention & Destruction Policy
Last updated: 30 August 2026. This is the public retention schedule some state biometric-privacy laws require. It covers every kind of voice-derived data Voxir handles.
What we hold
Numeric voice measurements (pitch, pitch variation, steadiness measures including jitter and shimmer, harmonics-to-noise ratio, loudness, speaking rate, pauses). Some laws may treat these as biometric information. We do not create voiceprints, do not build speaker models, and cannot identify anyone by their voice. Optionally — under a separate, written research consent (18+) — a small number of short research clips, the only audio that ever leaves a device.
Retention schedule
| Data | Kept | Destroyed |
|---|---|---|
| Voice measurements & score history | While your account is active | Immediately on account deletion or consent withdrawal |
| Transcripts | At most 30 days | Automatic purge; immediately on deletion or withdrawal |
| Research clips (optional, 18+) | For the duration of the study | At study end; immediately on withdrawal or deletion |
| Encrypted backups | At most 30 days | Age out automatically, in at most 30 days |
Measurements are kept while your account is active so you can see progress; transcripts are deleted within 30 days; everything is deleted immediately when you delete your account or withdraw consent, and our backups are kept no longer than 30 days. We never sell, lease or trade your voice data, never use it for advertising, do not create voiceprints and cannot identify you by your voice.
Destruction method
Deletion is a hard delete of every server-side row and file, confirmed by an audited job; research clip files are unlinked together with their database rows, and the end-of-study purge is audited. In no case is any voice-derived data held longer than one year after the purpose for collecting it ends.
Security
Voice-derived data is encrypted in transit and at rest, access is limited to the systems that serve you, and every administrative access is logged.