Duudlaga Flow Privacy Update 2026: Always-On Encryption & Permanent Deletion
Duudlaga Flow now encrypts all user content by default, never stores raw audio after processing, and permanently deletes data on request. Read the 2026 privacy update.
What changed in 2026
This year we made privacy the default, not an option. Every Duudlaga Flow user now gets the following protections automatically โ no setup, no extra code, no friction.
Always-on encryption
Every transcription, note, and uploaded file is now encrypted at rest by default. We removed the old privacy-code mode because requiring users to remember a code was a barrier. Now encryption just works.
- Each account gets its own derived encryption key.
- Plaintext columns in the database stay empty for encrypted rows.
- Content is decrypted only when it is returned to the authenticated owner.
- Uploaded audio uses envelope encryption: a unique file key wrapped by the account key.
No stored raw audio
After your speech is converted to text, the raw audio recording is no longer kept on our servers. Only the resulting text is retained, and it is encrypted.
This means the most sensitive input โ your actual voice โ has the shortest possible lifetime in our infrastructure.
Permanent deletion
We added two clear actions:
- Clear My Data โ delete all transcripts, notes, meetings, uploads, and generated audio while keeping your account.
- Delete Account โ full account deletion with all content removed.
In both cases, the content is erased. There is no restore, no hidden retention, no shadow copy.
Clearer consent on iOS
The iOS app now shows a scrollable data-sharing disclosure before any audio is sent. It names Google and OpenAI, explains what each receives, and requires the user to agree before dictation begins. This matches Appleโs latest privacy guidelines and gives users a clear understanding of what happens to their voice.
On-premise deployment for enterprises
Companies can now run the full Duudlaga Flow platform inside their own infrastructure. The only outside call is the speech-to-text service. Everything else โ storage, polish, formatting, admin dashboards โ stays on the local network.
The same two external providers
Nothing changed about who can touch your data:
| Provider | Receives | Purpose |
|---|---|---|
| Voice recording during dictation | Speech-to-text | |
| OpenAI | Transcribed text only | Polish, format, translate, AI text actions |
Audio never goes to OpenAI.
Why we did this
We want Duudlaga Flow to be the best Mongolian speech-to-text product available. But speed and accuracy do not matter if users cannot trust the tool with their words.
Your voice. Your thoughts. Your words. Yours.