Privacy Policy
Effective: March 2026
1. Data Controller
mazdek GmbH, Switzerland. Email: [email protected]
mazdek GmbH ("mazdek", "we", "us") is the controller of your personal data under the Swiss Federal Act on Data Protection (nDSG/FADP).
You must be at least 13 years old to create an account or use mazdek services.
2. Data We Collect
Account data — Email address, name (optional), profile picture (optional), authentication credentials, and account preferences.
Phone number (verification) — If account verification requires it, we collect and store your phone number to send a one-time SMS code and to check it against fraud/abuse signals. We use it only for verification and security.
Usage data — Transcription content and audio files you upload, chat conversations and messages, text-to-speech generations, songs you create (your lyrics, style descriptions and any photos you add as inspiration), documents and files you upload to projects, knowledge you create, translation requests, and project data.
Text voiced with platform voices — Some of our platform voices are based on recordings of real people who have licensed their voice to us or made it available under an open licence; the others are synthetically generated and have no real person behind them. When you convert text to speech with one of these platform voices, the text is checked for swear words, sexual content and hate speech before it is voiced: by a word list on our own servers and, unless you have declined third-party AI processing, by our content-moderation provider. If the text is refused, we keep a record of the attempt (see Data Retention). Audio generated with a platform voice contains a machine-readable "AI-generated" label and a provenance identifier that links the file to our record of the generation (your account, the voice, the time and an irreversible hash of the text — not the text itself). Political and news content is not filtered.
License record of generated audio — From 4 November 2026, when you generate audio on a surface where you provide the text (for example the text-to-speech page, Studio, the API, site integrations or Telegram), we keep a record of the generation for every voice: your account, the plan you held at that moment, the voice, the surface, the time, an irreversible hash of the text and an irreversible hash of the delivered file — never the text or the audio itself. The file carries the record's identifier. This lets us confirm which licence under our Terms (Section 4, commercial use of generated audio) applies to a given file. Reading chat answers aloud, Voice Mode, live interpreting, songs, clips and news bulletins are not recorded.
Voice samples & custom voices — If you create a custom voice, we collect the voice recordings you provide and use them to build a voice model. Voice samples are stored privately, and the voice model is created with our text-to-speech provider (see International Data Transfers). Voice is sensitive data; we process it only to provide this feature to you.
Technical data — IP address, browser type, device information, and session data. We use this to operate and secure the Service and to prevent fraud and abuse. From your IP address we derive an approximate location (country and city) for internal usage statistics; the IP address itself is not stored alongside those statistics, and the location is city-level at best — it does not identify your address.
Billing data — Subscription plan, usage statistics, payment transactions, and invoices. Payment card details are processed and stored exclusively by our PCI-compliant payment processor — we never see or store your card number.
Advertising and measurement identifiers — Only with your consent, we and our advertising partners process an advertising identifier (on Android, the device advertising ID; on iOS, Apple's privacy-preserving SKAdNetwork signals, which are aggregated and do not identify you) together with cookies or similar web technologies, solely to measure whether an advertisement led to an installation, a registration or a purchase. These identifiers are never used to read your content. You can reset or delete the advertising ID in your Android device settings, and you can change or withdraw your consent at any time in the cookie settings on the web.
Job applications — If you apply through our careers page, we process the details you enter and the CV you upload solely to assess your application. They are stored privately, are accessible only to the people who handle recruitment, and are deleted after six months unless you ask us to delete them earlier.
mazdek AI Keyboard — The optional mazdek AI Keyboard (iPhone keyboard extension, Android input method) works entirely on your device. We do not collect what you type: it is never sent to us or to anyone else, never stored on our servers and never used to train models. Suggestions and corrections are computed on the device from a dictionary that ships with the app. The words the keyboard learns from you (your personal dictionary) and its settings are stored only on your device; you can delete them at any time from the keyboard's menu or by removing the keyboard. The keyboard contains no network code and does not ask for "Full Access" on iPhone. In password and other sensitive fields it shows no suggestions and learns nothing.
3. Purposes and Legal Basis
| Purpose | Legal basis (nDSG) |
|---|---|
| Providing the service (ASR, TTS, chat, translation, song creation) | Contract performance |
| Account management and authentication | Contract performance |
| Billing and payment processing | Contract performance, legal obligation (OR 958f) |
| Security, fraud prevention, and abuse protection | Legitimate interest |
| Protecting the people behind platform voices (checking text before it is voiced, keeping evidence of refused attempts and provenance records) | Legitimate interest (protecting third parties' personality rights, establishing and defending legal claims) |
| Confirming which licence applies to generated audio (Terms, Section 4) and enforcing it | Legitimate interest (establishing, exercising and defending legal claims) |
| Service analytics and improvement (we do not train AI models on your content without consent) | Legitimate interest, with opt-out via privacy settings |
| Platform communications and feature updates | Legitimate interest |
| Advertising measurement and attribution | Consent, withdrawable at any time |
4. International Data Transfers
Your data is primarily processed on servers in the European Union. To provide the Service we rely on a small number of carefully selected service providers ("sub-processors"), described below by category. Some are located outside Switzerland, including in the EU and the United States:
| Category of recipient | Location | Purpose |
|---|---|---|
| Cloud hosting, database & file storage | EU | Running the Service and storing your data |
| GPU / AI compute providers | EU & USA | Speech recognition, text-to-speech, chat, language, image and music processing |
| Speech-recognition provider | USA | Transcription of uploaded audio and video files (receives the audio; not used to train its models) |
| Speech / voice processing provider | USA | Multilingual text-to-speech and custom voices (receives voice samples) |
| Payment processor | USA | Billing; card details are handled solely by the processor |
| SMS / communications provider | EU | One-time verification codes (receives phone number) |
| Fraud-prevention provider | USA | Risk scoring (receives IP address and, for verification, phone number) |
| Email provider | EU | Transactional email (verification, sign-in, notices) |
| Error-monitoring provider | USA | Diagnostics and reliability (no account identifiers attached) |
| Content-moderation provider | USA | Automatic check of posts, messages and text voiced with platform voices for harmful content (not used if you declined third-party AI processing) |
| Advertising and measurement platforms | USA | Campaign measurement and attribution (receive an advertising identifier and event data, only with consent) |
For transfers to the USA, we rely on the Swiss-U.S. Data Privacy Framework where applicable, or standard contractual clauses (SCCs) recognized by the FDPIC. A current list of our specific sub-processors is available on request at [email protected].
Our chat, translation and language-processing providers do not use your content to train their models. Our multilingual speech provider may use text and voice samples sent to it to improve its own speech models; only content you deliberately submit for multilingual speech synthesis or a custom voice reaches that provider.
5. Data Retention
| Data type | Retention |
|---|---|
| Account data | Preserved while deactivated; erased + anonymized on permanent deletion |
| Conversations, transcripts, TTS, songs, knowledge | Until you delete them individually |
| Audio files (uploads, generated audio) | Until you delete them individually |
| Videos uploaded for subtitles (Studio) | 7 days, then deleted automatically; you can delete earlier. Subtitle files remain until you delete the transcript. Dubbed versions (Max) are deleted together with the video |
| Billing records (invoices, subscriptions, transactions) | 10 years (OR Art. 958f) |
| Community feed (Civat) view records: which posts were shown to you, used only so the author's "seen by" count counts each person once | 30 days; the author sees only the total number, never who saw the post |
| Sessions | Removed on logout or account closure; expired sessions are not reusable |
| Background processing data (queues) | Auto-expires within days |
| Job applications sent through the careers page (form details and CV) | 6 months after submission, or earlier at your request |
| Abuse-prevention fingerprint kept on permanent deletion (irreversible keyed hash of email, phone and linked social identity — no readable data) | Until the end of the billing period in which the deletion happened |
| Platform-voice misuse evidence: refused text-to-speech attempts (account, voice, matched words, first 500 characters of the text) and provenance records of audio generated with platform voices (account, voice, time, irreversible hash of the text) (records created before 4 November 2026) | 12 months, also after account deletion |
| License record of generated audio (from 4 November 2026): account, plan at the time of generation, voice, surface, time and file hash; the text hash only for the first 12 months | 3 years from generation, also after account deletion; a record tied to a specific open dispute is kept until that dispute is resolved |
When you deactivate your account, your login is disabled but your data is preserved; you can reactivate by contacting support. When you permanently delete your data (Settings > Account > "Permanently delete my data"), your conversations, transcripts, generated audio, custom voices and personal profile are erased immediately and your profile is anonymized. Only financial records required by Swiss commercial law (OR Art. 958f, 10 years) are retained, in an anonymized form no longer linked to your identity. You can also request removal at [email protected].
6. Your Rights under Swiss Law
Under the nDSG (Art. 25–29), you have the right to:
- Access — Request a copy of your personal data (available via Settings > Privacy > Export Data)
- Rectification — Correct inaccurate personal data (via Settings > Account)
- Deactivation — Temporarily disable your account while your data is preserved (via Settings > Account)
- Erasure — Permanently delete your conversations, media and personal profile (via Settings > Account > "Permanently delete my data"). Billing records required by law are retained in anonymized form
- Data portability — Export your data in a structured, machine-readable format (JSON)
- Objection — Object to data processing for service improvement (toggle off "Help improve mazdek" in Privacy settings)
- Withdraw consent — Revoke consent for location metadata or AI training at any time via Privacy settings
To exercise any right not available through the app, contact us at [email protected]. We will respond within 30 days.
7. Privacy Controls
You can manage the following in Settings > Privacy:
- Help improve mazdek — Whether your de-identified queries may be retained to analyze and improve the Service. We do not use your content to train AI models without your consent.
- Export data — Download all your data as a JSON file at any time.
Both toggles default to off. Your preferences are saved to your account.
8. Data Security
We implement appropriate technical and organizational measures to protect your data (nDSG Art. 8), including:
- TLS encryption for all data in transit
- Encrypted database connections
- Session-based authentication with secure cookies
- Rate limiting and abuse protection
- Role-based access control for administrative functions
- Payment processing delegated to a PCI DSS-compliant payment provider
9. Supervisory Authority
If you believe your data protection rights have been violated, you may lodge a complaint with the Federal Data Protection and Information Commissioner (FDPIC):
FDPIC — Eidgenössischer Datenschutz- und Öffentlichkeitsbeauftragter, Feldeggweg 1, CH-3003 Bern — www.edoeb.admin.ch
10. Changes to This Policy
We may update this privacy policy from time to time. Material changes will be communicated via email or in-app notification. The "Effective" date at the top reflects the latest version.
11. Contact
For privacy-related questions or to exercise your rights: mazdek GmbH, Email: [email protected]
Last updated: November 2026