AI policy & security
Your IP stays yours. Your data stays private.
Studios work under their clients’ rules, and many clients have strict ones about AI. Here is what Mirva commits to, in plain terms: who owns the work, what happens to your data, where it lives, and what we sign.
The IP is yours
The IP and the artwork always belong to Mirva’s client.
No training on your work
Work on team and studio accounts is never used to train AI models.
Private, GDPR-compliant storage
Your data is kept in privately hosted storage, GDPR-compliant and under a Data Processing Agreement. EU hosting is available.
A mutual NDA, always
We sign a mutual NDA with every studio during onboarding, covering support and implementation as well.
For enterprise
Your own storage
Keep your data in your own private S3 buckets, under your own controls and retention.
Managed private cloud
Mirva deployed in a private cloud and run for you, by arrangement, for studios whose clients require it.
Choose where your data goes
Approved providers only
Every model in Mirva is run by a provider with its own terms. On request, we limit your team to the providers you or your client approve, and the rest are switched off for it.
A local model, on premises
With Mirva’s desktop app, the AI assistant can run on a local language model on your own hardware, on request, for work that must not leave the building.
Control inside the studio
Roles and permissions
Custom roles decide who can view, draw, edit documents and media, or see restricted projects.
Private and restricted
Boards can be private or closed, and projects restricted to the people on them. Guests see only what you share.
A record of who made what
Every stroke is recorded with who drew it, and every generation with its inputs and model, so you can show a client exactly how AI was used. Provenance →
When a client restricts AI
Many client policies come down to three questions: does the work stay ours, is it used to train AI, and can you show what the AI did? With Mirva the work stays yours, Mirva never trains on it, your team can be limited to the providers your client approves, and every step is on record. We’re glad to go through a client’s requirements with you, under NDA, before a project starts.
Questions
Who owns the art made in Mirva?
You do. The IP and the artwork always belong to Mirva’s client: your studio, or you.
Is our work used to train AI models?
No. Work on team and studio accounts is never used to train AI models.
Where is our data stored?
In privately hosted storage that is GDPR-compliant, under a Data Processing Agreement. EU hosting is available, and enterprise clients can keep their data in their own S3 buckets.
Can Mirva run in our own cloud?
Yes, as a managed private cloud deployment, by arrangement. Talk to us about your requirements.
Can the AI run on our own hardware?
Yes, on request: with Mirva’s desktop app, the AI assistant can run on a local language model on your own machines, so its conversations never leave your premises.
Do you sign an NDA?
Always. We sign a mutual NDA with every studio during onboarding, and it covers support and implementation too.
Which AI providers see our data?
A generation’s inputs go to the model that runs it. During onboarding, under the NDA, we go through each model you plan to use and its provider’s terms. On request, we limit your team to the providers you or your client approve.
This page summarizes Mirva’s commitments to studios. The binding terms are in your agreement with Mirva, its DPA and the mutual NDA.