1. The short version
Eligible everyday dictation is processed on your iPhone. Dictaphone does not receive that audio or nearby text. Meeting recording is different: after you acknowledge that every participant consented and finish recording, the meeting audio is sent over an encrypted connection to the Dictaphone API and transcribed only with Groq.
2. Data processed on your device
On-device dictation, app-aware field classification, deterministic text tools, anonymous meeting speaker clustering, your personal dictionary and saved transcripts can be processed locally. Saved transcripts remain under your control on the device unless you use an Apple system feature that synchronises app data.
3. Cloud meeting transcription
Meetings do not use Apple’s local transcriber. Dictaphone uploads temporary, encrypted audio chunks to our API, which forwards them to Groq’s Whisper Large v3 Turbo speech model. Preferred spellings from your personal dictionary and a short excerpt from the preceding transcript may accompany each chunk to improve spelling and continuity. We configure the production provider for zero data retention. Our API returns the transcript to the app and does not intentionally store recording, dictionary or transcript content.
The app may keep protected temporary audio locally for up to 15 minutes if processing fails so you can retry. It is deleted after successful processing, when the retry window expires, or when you delete the meeting.
4. Participant consent
You must inform every participant that recording and cloud AI transcription will occur and obtain any consent required by law, contract or workplace policy. The app requires an acknowledgement before recording starts, but Dictaphone cannot verify that another person actually consented.
5. Metadata we keep
We retain operational metadata needed to run and protect the service: pseudonymous workspace identifiers, App Store product and entitlement status, transcription time, billed time, provider/model identifiers, estimated provider cost, request identifiers, timestamps, privacy settings, and security or idempotency records. These records do not contain the transcript or raw audio.
6. Apple subscriptions
Purchases are made through Apple. Dictaphone receives signed transaction and entitlement information such as product ID, expiry, environment and transaction identifiers. Consumer access does not require you to give Dictaphone an email address.
7. Providers and international processing
Groq processes meeting audio for speech transcription. Railway hosts the Dictaphone API and database. Apple provides subscription and device services. Depending on your location, this can involve processing outside your country, including in the United States, with the safeguards offered by those providers and our contractual arrangements.
8. Retention and deletion
Cloud recording and transcript content is handled transiently. Operational metadata is retained while needed for billing, fraud prevention, security, service operation and legal obligations. You can delete local transcripts and meeting artifacts in the app. Contact us to request deletion of server-side account metadata where applicable.
9. Your choices
You can use eligible on-device dictation without using cloud meeting hours, choose whether to save a meeting transcript locally, stop or delete a recording, decline meeting consent, and manage or cancel subscriptions through Apple.
10. Security and changes
We use transport encryption, short-lived app access tokens, protected local files and content-blind administration. No system is risk-free. We may update this policy when the product or law changes and will publish the effective date here.
11. Contact
Privacy questions and requests can be sent to venkat@amzuit.com.