AI & Data Use
Last updated: September 30, 2026
What each feature sends
| Feature | What leaves your computer | Where it goes |
|---|---|---|
| Sign-in, plan, credit balance | Email, password over TLS, app version | Mozonic (AWS, United States) |
| AI assistant | Text only: your message plus session context — track names, tempo, key, markers, measurements, plugin names, notes you stored. No audio. | The model provider you pick |
| Cloud mix analysis | A rendered audio file of your mix | Mozonic storage, analyzed by Mozonic |
| Cloud audio generation | Your prompt and musical context; for audio-to-audio, a rendered section of your session | Google (Lyria) or Stability AI |
| Voice dictation | Nothing — speech is transcribed on your computer | The model downloads once from a CDN |
| Diagnostic logs | Timestamps, levels, messages, an install identifier, your account email and plan, OS and app versions, engine warnings. No audio, no project contents. | Datadog (United States) |
| Payments | Handled entirely by Stripe — we never receive your card details | Stripe |
| Plugin scanning | Nothing — your plugin list stays on disk | — |
Two things Mozonic deliberately does not do: it does not read iLok, PACE or any other plugin licensing system, and it does not upload your audio in order to improve anything.
Model providers and training
Requests go to providers under their commercial API terms, not their consumer products:
- •Anthropic (Claude) — does not train on API inputs or outputs. Content may be retained briefly for safety and abuse monitoring.
- •OpenAI (GPT) — does not train on API inputs or outputs. Content may be retained up to 30 days for abuse monitoring.
- •Google (Gemini, Lyria) — Mozonic uses the paid Gemini API tier, on which prompts and responses are not used to improve Google’s products.
- •Stability AI (Stable Audio) — governed by Stability’s own API terms, which you should read if this matters to your work.
Providers can change their terms. If one we use starts training on customer traffic, that model is removed from the list or this page is updated before it stays available.
Who owns what the AI produces
- •Your projects, recordings and exports are yours. Mozonic claims no ownership of them.
- •Generated audio is yours to use, subject to the generating provider’s terms — with two honest limits: generative models can produce output similar to output given to someone else, and Mozonic makes no representation that generated audio is free of third-party rights. If a release depends on clearance, clear it.
- •You are responsible for the rights in anything you upload or ask the agent to work on.
How long things are kept
- •Mixdowns sent from Mozonic Studio for cloud analysis: 7 days. The analyzer reads the audio once and writes its report to your account; nothing re-reads the audio afterwards, so it is deleted automatically.
- •Audio you upload to the web app — mixes, stems, ZIP sessions, and the corrected mixes, bounces and masters made from them — is kept so that you can come back to your reports and re-download your files. It stays until you delete it, or until you delete your account.
- •Diagnostic logs: 14 days.
- •Account, subscription and credit history: for the life of the account.
- •Model-provider retention: as described above, per provider.
- •Anything on your own machine: until you delete it.
Consent, opting out, and deletion
- •Cloud analysis and generation ask every time, naming the exact file size and duration before anything is uploaded. Declining uploads nothing and costs nothing.
- •The AI assistant sends session context only when you send a message.
- •Diagnostic logs are on by default in released desktop builds, so that a bug you report can be diagnosed. Ask support to turn them off for your account if you would rather they were.
- •Deleting your account removes your server-side data, cancels billing and purges uploaded audio. It does not delete anything on your own computer — your projects, exports and local logs stay where they are, and stay openable, because Mozonic Studio never requires an account to open your own work.
Who processes data for us
Amazon Web Services (hosting and storage), MongoDB Atlas (database), Stripe (payments), Datadog (diagnostics), Resend (email), Sanity (website content), Anthropic, OpenAI, Google and Stability AI (AI models), and a public CDN for the one-time download of the on-device speech model. All processing is in the United States.