Is AI try-on safe? What actually happens to your photo

Every virtual try-on app needs a photo of you. That is not a design choice anyone can engineer around — the entire proposition is seeing a garment on your body, and there is no way to do that without your body.

So the question is not whether to hand over a photo. It is what happens to it afterwards. Here are the four questions worth asking any app in this category, why each one matters, and how Utopia answers them.

1. What is the photo used for?

The specific thing to look for is whether the photo has one purpose or an open-ended list of them. A privacy policy that says a photo is used 'to provide and improve our services, for research, and for other purposes described herein' is not describing a single use — it is reserving the right to decide later.

Utopia: your photo builds your fit model — the reusable likeness that every try-on is generated against — and nothing else. There is no second purpose. The mechanism is described in full in how virtual try-on works, and it is worth understanding precisely because it makes the scope of the use auditable.

2. Who can see what you generate?

The failure mode people worry about is a public feed with a default-on setting. It is a reasonable worry: an app whose growth depends on shared content has an incentive to make sharing the default and the opt-out quiet.

The thing to check is what happens if you never touch a single setting. If the honest answer is 'other people can see your images', that is a default-public product regardless of what the settings screen offers.

Utopia: nothing you generate is public unless you choose to share it. Sharing is an action you take on a specific look — it is not a mode you are in. You choose per look whether it is private, visible to friends, or public.

3. Does my photo train the model?

This is the question with the most weight and the least clarity across the industry. 'We do not sell your data' is not an answer to it. Neither is 'we may use content to improve our services', which is the standard phrasing for exactly the thing you were asking about.

It matters because training is effectively irreversible in a way that storage is not. A stored file can be deleted; a model that has already learned from it cannot straightforwardly unlearn.

Utopia: your photos are yours. They are not used to train a generative model. The current privacy policy states what is collected, what it is used for and what is retained, in the specific terms Apple's App Review requires for face data — and it is worth reading rather than taking a blog post's word for it.

4. Is deletion actually deletion?

Two different things get called deletion. One removes an item from your view while retaining it on a server. The other removes it. Only the second is worth anything, and the distinction is usually made in a sentence you have to go looking for.

The related question is whether account deletion is possible from inside the app at all. Apple has required this of apps that support account creation for some years, but the quality of the implementation varies enormously.

Utopia: deleting your account permanently removes your photos, your fit model, your try-ons and your looks. It is available in the app — Profile ▸ ••• — and the support page documents the route.

The question that gets missed: other people

There is a fifth question, and it is the one most discussions of AI photo safety skip entirely: what stops someone using this on a photo of you, without you?

Any app that can put a garment on a person's photo can, in principle, put a garment on a person's photo they did not consent to. The honest position is that this is a policy and enforcement problem rather than a technical one, and an app that does not acknowledge it has not thought about it.

Utopia: the product is explicitly built to let you style other people — a friend, a partner, a parent — because that is genuinely one of the best things about it. The corresponding obligation is enforcement: non-consensual imagery is removed on sight, reporting is available on every shared look, and the terms set out what is not allowed.

A checklist you can use on any app

1. 01Does the privacy policy name one purpose for your photo, or reserve a category of purposes?
2. 02What is visible to others if you change no settings at all?
3. 03Does it say plainly whether your images train a model — in those words?
4. 04Does account deletion remove the underlying files, and can you do it from inside the app?
5. 05Is there a stated policy and a working report button for images of people who did not consent?

If an app cannot answer all five in language you understand without a lawyer, that is itself the answer.

Read the policy first if you like — it is at https://getutopia.app/privacy, in plain English.

Get the app