ClaritateAI is a voice-memo recording and transcription app by Bronn Software (Geran Bronn). This policy explains what data the app handles, where it goes, how long it is kept, and the choices you have. The short version, which we consider the core promise of the product:
Nothing you upload stays on our servers. Recordings live on your device. When you ask for a transcription, the audio is uploaded, transcribed, and deleted from our servers the moment transcription completes. The same applies to everything else you send us: a video is reduced to its audio track and the video is deleted immediately; a PDF or an image is deleted the moment its text has been read. Only the text remains, until you delete it.
Important: ClaritateAI uses third-party AI services
ClaritateAI cannot transcribe or summarise anything on your device alone. To do those things, the content you choose to process is sent to third-party artificial-intelligence providers outside our company, listed by name in section 4 below.
In plain terms:
- Your audio recordings are sent to an AI transcription provider to be turned into text.
- Your transcripts and the text of documents you send are sent to an AI language provider when — and only when — you use a summary, translation, checklist or "Ask AI" feature.
- Images, and pages of scanned PDFs, are sent to an AI vision provider to be read.
- Your audio is additionally sent to a speaker-identification provider if you switch on "Identify speakers".
Nothing is sent anywhere until you ask for it. The app asks for your explicit permission before the first time any of your content leaves your device, and you can withdraw that permission at any time (see section 6). Recording, playback and everything else stored locally continue to work if you say no; the AI features do not.
These providers process your content only to return the result you asked for. Your content is not used to train AI models, and it is not used for advertising or sold to anyone.
1. Information we collect
1.1 Account information
When you sign in with Google (or Apple, on supported devices), we receive and store:
- Your name and email address
- Your profile picture URL (if any)
- A unique account identifier from the sign-in provider
We use this to create your account, sync your transcripts across sessions, send you service emails, and manage your subscription plan. We never see or store your Google/Apple password.
1.2 Audio recordings
- Recordings are made and stored on your device. Nothing is uploaded until you request a transcription (or enable auto-transcribe in Settings).
- When you request transcription, the audio file is uploaded to our server over an encrypted connection (HTTPS), processed, and then immediately and permanently deleted from our server once the transcription succeeds.
- If a transcription fails, the uploaded audio is kept temporarily so it can be retried, and is automatically deleted after 7 days at the latest.
- Deleting a recording in the app deletes it from your device. Because the server copy is already gone after transcription, a deleted recording's audio cannot be recovered by anyone, including us.
1.3 Video, documents and images you choose to send
You can also hand ClaritateAI a video file, a PDF, or a photo. Nothing is uploaded unless you pick it yourself, and none of it is kept:
- Video. Uploaded over HTTPS, then the audio track is separated out and the video is deleted from our server immediately — before the transcription itself even begins. We do not transcode, watch, index or retain the picture. Only the audio track continues through the normal transcription flow above, and it is deleted when transcription completes.
- PDFs. Uploaded, the text is extracted on our own server, and the file is deleted as soon as the text is out. If a PDF has no text layer (a scan, or photographed pages), its pages are rendered to images and read by a third-party AI vision provider (see section 4) so that they can be turned into text — the rendered images exist only for the duration of that request and are never stored.
- Images. Uploaded, read by a third-party AI vision provider (see section 4) so their contents can be turned into text, and deleted as soon as the reading returns.
- If any of these fails, the file is deleted within 24 hours at the latest rather than the 7 days that applies to audio — there is nothing to retry it from that your device does not still have.
We do not use anything you send us to train AI models, and neither do our providers under the terms we use them on.
1.4 Transcripts and AI content
The following text data is stored on our servers so it can be restored and synced, until you delete the item it belongs to:
- Transcripts (including timestamps and speaker labels)
- Text extracted from documents and images
- Speaker names you assign
- AI summaries (in each style you generate), AI translations, and extracted checklists
- Your edits to transcript text
Your "Ask AI" chat history, and which checklist items you have ticked off, are stored on your device only.
1.5 Usage records
For quota and billing purposes we record, per transcription and AI action: minutes of audio processed or pages read, the feature used, and a cost estimate. These records do not contain audio, document or transcript content and are retained for accounting purposes.
1.6 What we do NOT collect
- No advertising identifiers, no ad networks, no data sale — ever.
- No third-party analytics or tracking SDKs.
- No contacts, location, or browsing of your photo library or files. We only ever receive the specific file you pick or share into the app.
2. Permissions the app uses
| Permission | Why |
|---|---|
| Microphone | Recording voice memos — the app's core function |
| Notifications (optional) | Status notifications; you can decline |
| (none for files) | Picking a video, PDF or image uses your phone's own file and photo pickers, which hand us only the one file you choose. The app never asks for access to your photo library or storage. |
| Foreground service | Keeps a recording or upload running when the app is in the background; always shows a visible notification while active |
3. What we send to AI providers, and when
Every transfer below happens only in response to an action you take. Nothing is sent in the background, on a schedule, or "just in case".
| When you… | What leaves your device | Where it goes |
|---|---|---|
| Ask for a transcription (or switch on auto-transcribe) | The audio recording, and the filename you gave it | Our server, then an AI transcription provider |
| Send a video | The video file; we strip the audio track and delete the video before transcribing | Our server; only the audio track reaches the AI provider |
| Switch on "Identify speakers" | The same audio, additionally | A speaker-identification provider |
| Send a PDF with a text layer | The PDF | Our server only — no AI provider is involved |
| Send a scanned PDF or an image | The image, or the scanned pages rendered as images | Our server, then an AI vision provider |
| Use a summary, translation, checklist, or "Ask AI" | The transcript text (or the text extracted from your document), plus the question or instruction you typed | Our server, then an AI language provider |
| Nothing — you only record and play back | Nothing | — |
What we never send to an AI provider: your name, your email address, your account identifier, your payment details, or your device identifiers. The providers receive the content to be processed and nothing that identifies you as its author.
Auto-transcribe, if you enable it in Settings, sends each finished recording automatically. It is off by default, and switching it on is a standing instruction to transfer — you can withdraw it at any time by switching it back off.
4. Third-party AI providers and other processors
These are the companies that process your content on our behalf. Each acts strictly on the operation requested, never for advertising, and — as set out in the terms we use them on — does not use your content to train models.
| Provider | Operated from | Purpose | What it receives | Retention |
|---|---|---|---|---|
| DeepInfra, Inc. | United States | Speech-to-text transcription (our primary provider) | The audio file, for the duration of processing | Not retained after the request; not used for training |
| DeepInfra, Inc. | United States | Reading images, and pages of PDFs with no text layer | The image, or the rendered page, for the duration of the request | Not retained after the request; not used for training |
| OpenAI, L.L.C., Groq, Inc., or OpenRouter, Inc. | United States | Standby transcription providers, used only if our primary provider is unavailable | The audio file, for the duration of processing | Per each provider's API terms; API content is not used for training |
| pyannoteAI | France (European Union) | Speaker identification — only if you switch on "Identify speakers" | The audio file, held in temporary storage while it is analysed | Deleted automatically, typically within 24–48 hours |
| DeepSeek | People's Republic of China | AI summaries, "Ask AI" questions, translations and checklist extraction — only when you use one of those features | The transcript or document text, and the question or instruction you typed. Never your audio, and never your images. | Per their API terms |
| Mailgun (EU region) | European Union | "Ready to read" service emails | Your email address, the item's name, and its length or page count | Standard email delivery logs |
Please read this if you handle confidential material. Our AI language provider, DeepSeek, operates from the People's Republic of China, and text you send to a summary, translation, checklist or "Ask AI" feature is processed on servers in that jurisdiction, where the legal protections and government access powers differ from those in your own country. This affects the text features only — your audio recordings are never sent to DeepSeek. If your transcripts contain material you are not willing to have processed under those conditions, simply do not use the summary, translation, checklist or "Ask AI" features; transcription works without them.
Text extraction from PDFs that already contain a text layer happens on our own servers and involves no third party at all.
Our own servers are operated on infrastructure rented from a hosting provider. All traffic between the app, our servers and these processors is encrypted in transit (HTTPS/TLS).
International transfers
Using ClaritateAI's AI features involves transferring your content outside your own country, and — depending on where you are — outside your region's data-protection area. Where the law requires a transfer safeguard, we rely on the processor's standard contractual clauses and, for transfers you initiate to a country without an adequacy decision, on your explicit consent to that transfer (GDPR Art. 49(1)(a); the corresponding provisions of POPIA s.72 for South African users). Section 6 explains how that consent is obtained and withdrawn.
5. How long we keep data
| Data | Kept until |
|---|---|
| Audio on our servers | Deleted the moment transcription completes (7 days max for failed jobs) |
| Video on our servers | Deleted the moment its audio track has been separated out, before transcription begins (24 hours max if that fails) |
| PDFs and images on our servers | Deleted the moment their text has been read (24 hours max if that fails) |
| Audio on your device | Until you delete the recording |
| Transcripts, extracted text, summaries, translations, checklists, speaker names | Until you delete the item, or your account is deleted |
| Account details | Until your account is deleted |
| Usage/billing records | Retained as required for accounting |
| Anti-abuse quota record on account deletion (a one-way hashed identifier and your remaining-minutes counter — no name, email, or content) | Until the billing cycle that was in progress at deletion ends, then deleted automatically |
6. Your consent
Because using ClaritateAI's AI features means your content is shared with the third-party providers named in section 4, we ask for your permission before that happens rather than assuming it.
How we ask. The first time you do something that would send your content off your device, the app shows you a disclosure screen naming the data involved and the providers that will receive it, and asks you to agree. Nothing is transmitted until you do. This is separate from accepting this policy: agreeing to the policy alone is not treated as consent to AI processing.
What happens if you decline. The app keeps working as a recorder. You can record, play back, rename, organise into folders and delete — all of it on your device. Transcription and the AI features stay unavailable until you consent, and you will be asked again only if you try to use one of them.
Withdrawing consent. You can withdraw at any time in Settings → Privacy → AI processing. From that moment nothing further is sent to any AI provider. Withdrawal is not retroactive: it cannot recall content already processed, and transcripts you have already generated remain in your account until you delete them.
Our legal basis for this processing is your consent (GDPR Art. 6(1)(a); POPIA s.11(1)(a)). Where a transcript happens to contain special-category or sensitive personal information, our basis is your explicit consent (GDPR Art. 9(2)(a); POPIA s.27(1)(a)) — which is one reason we ask separately rather than bundling it into acceptance of this policy.
A note on other people's voices. A recording often captures people other than you. You are responsible for having whatever permission your local law requires before recording and transcribing someone else — in some countries and some states every participant must consent to being recorded. Please check what applies where you are.
7. Your rights and choices
- Delete an item: removes the device copy and all associated server-side text data (transcript or extracted text, summaries, translation, checklist, speakers).
- Delete your account: use Settings → Delete account in the app, or contact us at admin@bronnsoftware.co.za. This deletes your account and all associated data. To prevent abuse of free-tier limits, a one-way hashed identifier and your quota counters are kept until the billing cycle in progress at deletion ends — this cannot be linked back to you and is deleted automatically. See the account deletion page for step-by-step instructions (including how to delete without the app installed).
- Access/export: transcripts can be exported from the app at any time (TXT, PDF, SRT); you may also request a copy of your stored data by email.
- Decline or withdraw AI processing: see section 6. Speaker identification, the AI text features, and notifications are each additionally optional and off unless you turn them on.
If you are in the European Economic Area or the United Kingdom, you additionally have the rights granted by the GDPR (access, rectification, erasure, restriction, portability, objection, and withdrawal of consent). If you are in South Africa, you have the equivalent rights under POPIA. To exercise any of these, email admin@bronnsoftware.co.za.
8. Security
Data in transit is encrypted with TLS. Sessions use expiring tokens. Server-side audio exists only transiently (see section 1.2). No system is perfectly secure, but our architecture is deliberately designed so that the most sensitive data — your voice — spends the least possible time anywhere other than your device.
9. Children
ClaritateAI is not directed at children under 13 (or the equivalent minimum age in your jurisdiction), and we do not knowingly collect data from them.
10. Changes to this policy
We may update this policy as the product evolves. Material changes will be announced in the app or by email, and the effective date above will be updated. Continued use after changes take effect constitutes acceptance.
If we ever add a new AI provider, or send a category of your data somewhere it has not gone before, we will ask for your consent again in the app rather than relying on a changed policy page you may never see.
11. Contact
For any privacy question, data request, or complaint, email us and we will respond as soon as reasonably possible.
Bronn Software
South Africa
admin@bronnsoftware.co.za