AfterBell beta privacy guidance ยท version M3-2026-08-19
Privacy and data
AfterBell is designed to reduce unnecessary exposure, not to promise that every identifier will always be detected.
What happens to a voice debrief
Your browser records audio and sends it temporarily over HTTPS to the trusted AfterBell server for Faster-Whisper transcription. The transcript is returned to you for review and editing. It is sent for further processing only when you explicitly choose Send.
AfterBell then removes or replaces detected direct identifiers before sending anonymised content to the configured external LLM provider. Some anonymised content therefore leaves AfterBell. The external provider's terms and data handling apply to that anonymised request.
What is not retained as durable AfterBell data
Raw audio, raw transcript, identity mappings, rehydrated teacher-visible text, and generated speech audio are temporary processing data. They are not written to the durable AfterBell record store, logs, or backups as product records.
Anonymised debrief records may be retained on the trusted AfterBell server for up to 30 days for the beta. Encrypted recovery backups are a separate rolling copy and may retain deleted live data until their seven-day expiry. AfterBell does not claim instant erasure from backups.
Important limitation
Anonymisation can miss an identifier, especially when names, places, or other details are unusual or transcribed imperfectly. Do not submit content that requires a guarantee of complete anonymity.