Privacy Policy
Last updated: August 31, 2026
This policy explains what Zonic AI does with the information and media you give it: where copies live, who else processes them, how long they are kept, and what we will never do with them.
Who this covers
Two different groups of people appear in this policy. The first is you, our customer — the account holder who uploads footage and runs edits. The second is everyone who appears or is heard in that footage, above all interview subjects, who never signed up for anything.
For your account data we are the controller. For the footage you upload we act as a processor on your instructions: we handle it to run the service you asked for, and for nothing else.
What we process
Three kinds of information:
- Account data — your name, email address, and profile picture from Google sign-in. We never receive or store a password; sign-in is delegated to Google.
- Media and everything derived from it — the video, audio, images, and documents you upload, plus the transcripts, translations, scene descriptions, embeddings, face and speaker groupings, edit plans, and chat history the service generates from them.
- Operational data — diagnostics and logs needed to run, debug, and secure the service, and billing records for paid plans.
People who appear in your footage
Interview footage contains other people's personal data: their face, their voice, their opinions, and often their name and employer. Zonic derives more from it — a transcript of what they said, and groupings that recognise the same face or voice across clips.
We process that material only to run your project. We do not build a cross-customer identity database: face and speaker groupings are created inside a single project and are never shared, matched, or reused across projects or customers.
The lawful basis for filming those people and for having us process the footage is yours, not ours. If an interview subject withdraws consent, tell us and we will delete the material from your project.
What we use it for
To run the service you asked for: analysing footage, transcribing and translating speech, making it searchable, generating and revising edit plans, rendering and exporting, answering support requests, keeping the service secure, and billing paid plans.
We do not sell personal information, and we do not use your footage or its transcripts for advertising or profiling.
Model training — what we do and do not do
We do not train, fine-tune, or otherwise adjust model weights on your footage. No such capability exists in Zonic: there is no fine-tuning job, no training corpus, and no dataset assembled from customer media.
What does improve over time is a library of written editorial rules — plain-language guidance such as "establish the location before the action". These are authored by our engineers as text and reviewed in our source repository like any other code. When we study a project in order to write one, we look at derived text — the edit plan, the chat, the scene descriptions — and never at the footage itself. Our own authoring rules forbid encoding anything specific to one customer, event, or project into them. None of this is automatic: it happens only when a person deliberately runs an internal tool.
Your footage, audio, interview content, and the clips cut from them are not placed into another customer's project or another project's output, and we do not provide or license them to any third party for that party's own purposes. The only outside parties who touch them are the processors listed below, acting on our instructions to run your project.
Where copies live
In cloud mode, your media is uploaded to our Google Cloud Storage bucket in Taiwan (asia-east1) and stays there until the project is deleted. Derived material — transcripts, scene descriptions, embeddings, face and speaker groupings, and chat — is stored with our managed database and cache providers in the United States, in every mode.
In local mode, your original files stay on your own machine and are not uploaded; Zonic reads them where they sit. Local mode is not an offline mode, and it is important not to mistake it for one: the analysis stages still send the footage to the AI providers listed below, derived text still goes to our cloud, and if you choose to share a project or submit a support log, that upload goes to our cloud too.
If you give us access to material in your own cloud drive, we copy what we need in order to process it — analysis cannot happen without reading the bytes. Those copies live in the project storage described above and are removed when the project is deleted.
When an engagement ends, tell us and we will delete the project. What deletion does and does not reach is described below, and we would rather you knew the limits in advance than discovered them later.
Who else processes your material
Running the service means passing material to infrastructure and AI providers. Each acts on our instructions, under its own contract with us, and none is licensed to use your material for its own purposes:
- No component of Zonic runs in mainland China, Hong Kong, or Macau, and no provider above is owned by a company headquartered there. The one Hong Kong item is a Stripe billing entity, which handles payment data and never touches media.
| Provider | What it receives | Where |
|---|---|---|
| Google Cloud (Storage, Cloud Run) | Stores project media; runs analysis and the agent | Taiwan (asia-east1) |
| Google Gemini API | Video, sampled frames, and text for scene understanding, subtitle cleanup, and translation | Operated by Google |
| Microsoft Azure AI Speech | The extracted audio track, for transcription | Operated by Microsoft |
| TwelveLabs | Video, when selected for scene analysis, search, or transcription | United States |
| OpenAI | Scene description and transcript text, for search embeddings. No media. | United States |
| Anthropic Claude, via Google Vertex AI | Chat prompts, transcripts, scene text, and sampled video frames | Not pinned to one region |
| Upstash (cache and vector search) | Transcripts, scene text, embeddings, chat, face and speaker groupings | United States |
| Vercel | Hosts the web application and brokers requests | Global edge network |
| Prisma / managed Postgres | Account and project records. No media. | Operated by the vendor |
| Stripe | Billing and payment data only. No media. | United States and Hong Kong entities |
| OpenStreetMap Nominatim | GPS coordinates read from footage, to name a location. Local mode only. | OpenStreetMap Foundation |
| Vimeo, YouTube, Google Drive | A finished video, only when you choose to publish it there | Operated by the vendor |
How long we keep it, and what deletion reaches
We keep project material until you delete it. Nothing expires automatically and nothing is removed behind your back. You can delete a project at any time from the app or through the API, which removes the media, the derived analysis, the search vectors, and the chat.
Three limits we would rather state plainly than let you assume otherwise. Deleting a single video removes the original but can leave generated derivatives — a preview, a thumbnail, a teaser — until the whole project is deleted. Deletion is immediate and final: we do not keep object versions or a soft-delete window, so there is no undo. And deleting a Zonic project does not currently send a deletion request to every AI provider that processed the material, so a copy may persist with a provider under that provider's own retention terms.
If your engagement requires verified downstream deletion, a written deletion certificate, or a fixed deletion deadline after termination, tell us before you upload. Those are commitments we can make for a specific customer, and they are described in the Security and Data Handling statement.
Account and billing records are kept while your account is open and for as long as tax and accounting law requires afterwards.
Your rights, and getting your data out
You can ask us for a copy of the personal data we hold about you, correct it, or have it deleted. There is no lock-in on your work either: your originals stay downloadable, project data is retrievable through the Platform API, and edit decisions export to standard editing project files rather than a format only we can read.
If you appear in footage but are not our customer, the organisation that filmed you controls that project and is the right first contact. If you cannot reach them, write to us and we will pass the request on and help where we can.
Depending on where you live you may have additional statutory rights. Write to us and we will honour them.
Contact
Privacy questions and data requests: support@zonic.tech