Legal — 11
AI Disclosure
Where the studio uses AI, what it can see, what it can never do, and how we use it in the work itself. Specifics rather than reassurances.
Last updated: 8 September 2026
- Studio
- Silvanè Studios
- Based in
- Ontario, Canada
- Contact
- hello@silvanestudios.com
1. Where AI appears
In one product feature: the studio concierge, which answers questions from signed-in clients about their own project — where it has got to, what we are waiting on, what has been approved, what has been paid. It is powered by a Google Gemini model called through Google’s API.
There is no chatbot on the public site. Nothing you read on the marketing pages is generated live, and no visitor data is fed to a model.
2. What the concierge can and cannot do
- It is read-only. It can answer questions. It cannot approve anything, change a project, move a date, spend money, or send a message on your behalf.
- It only ever sees your own organisation’s records. Every lookup is filtered by your memberships in the database query itself, so another client’s data is never fetched, let alone shown to a model.
- The model cannot choose whose data to read.None of the tools available to it accept an organisation identifier. The only identity in play is the session the server has already verified, so there is no instruction anyone could type that would make it fetch somebody else’s project.
- Studio staff are excluded from it.Asking the concierge is not a shortcut into every client’s data; that is a separate administrative surface with its own audit trail.
3. What is sent to the model
Your question, and the specific project facts needed to answer it — stage, dates, outstanding items, approvals, invoice status. Not your password, not payment details, not other clients’ information, and not the contents of files you have uploaded unless your question is about them.
Google states that data submitted through the paid Gemini API is not used to train its models. We rely on that commitment; it is Google’s to keep, and we link the vendor plainly on the Sub-processors page so you can read their terms yourself.
4. It can be wrong
Language models produce confident sentences whether or not they are correct. The concierge answers from your real project records, which removes most of the risk, but it is not a source of truth and nothing it says is a commitment by the studio. Anything that matters — a date, a price, an approval — is confirmed by a person. If its answer and your proposal disagree, the proposal is right and we want to hear about it.
You can always talk to a human instead. There is no phone tree here, and asking a person is not a downgrade.
5. AI in the work we produce
We use AI assistance in drafting and development, the way a studio uses any other tool. What that does not change:
- A person at Silvanè Studios reviews and edits everything before it goes out, and is accountable for it.
- Claims, figures, and legal or tax statements in published writing are checked against primary sources, which are cited.
- Design direction and code architecture are decided by people who can explain why.
- We do not publish generated imagery presented as photography of real people, places, or work.
If you would prefer no AI assistance anywhere in your project, say so at proposal stage and we will scope it that way.
6. Your content
We do not sell client content to model providers, and we do not use one client’s material to produce another’s work.
7. Changes
If we add an AI feature, change model provider, or extend what the concierge is allowed to do, this page changes first — before the feature ships, not after. Questions: hello@silvanestudios.com.