Official documentation
Diction privacy policy
Diction Labs, 2026
Read the originalRead on October 20, 2026
- What it is
- The Diction privacy policy, last updated October 2026. It describes how audio and text are handled in on-device, self-hosted and Diction One cloud modes, and what the app collects.
- How much weight it carries
- The maker's own privacy policy, so it describes the maker's design and claims. It does not name the model providers behind Diction One.
Facts taken from this source
- Diction's privacy policy says audio sent to Diction One is not saved and is discarded once transcription completes.
“Audio is never saved. Your audio is processed in memory and discarded the moment transcription completes.” (Section 'Diction One (Cloud)')
- Diction's privacy policy says its audio and transcriptions are never used to train, fine-tune or improve any models.
“Your audio and transcriptions are never used to train, fine-tune, or improve any models.” (Section 'Diction One (Cloud)', 'No model training.')
- Diction's privacy policy says on-device mode processes audio entirely on the iPhone with a local speech model.
“Audio is processed entirely on your iPhone using a local speech model.” (Section 'How Each Mode Handles Your Audio', On-Device Mode)