When you photograph a jumper for a wardrobe app, three things happen. The image goes onto your device. It gets sent to a server. Somewhere between the two, decisions get made about what happens to it. Most people do not think about these decisions. Most apps count on that.
What most wardrobe apps do with your photos
There is no single answer, because it depends on the app, but the common patterns are:
- The photo is uploaded to the company's storage, usually on Amazon Web Services or Google Cloud.
- A copy is passed through an image-classification model to work out what it is (jumper, dress, shirt, etc.).
- In many cases, the photo is added to a training set that improves the model over time.
- In some cases, the photo is also fed into third-party AI services, some of which retain a copy.
- The photo often has a public URL, even if it's a long random one, meaning anyone with that link can view it.
None of these are unusual for consumer apps. Most terms of service quietly cover them. That does not make them fine; it makes them worth understanding.
The parts that matter
Whether the photos train the model
If your wardrobe app improves its AI by training on user photos, your images are in the training set. This is the difference between an app that uses your photos to serve you and an app that also uses your photos to build a product. The line is subtle and important.
Look for a specific phrase in the privacy policy: "we do not use your images to train our models", or a version of it. Anything vaguer probably means yes.
Whether the storage URLs are public
A photo can live in a private bucket that only your account can see, or in a public bucket with a hard-to-guess URL. The second is easier to build, faster to serve, and slightly less private. Anyone who ever had that URL, including a third-party service passed the URL for processing, retains the ability to fetch the photo unless the URL is invalidated.
Look for the words "scoped", "per-user", "signed URL", or "short-lived URL". The good version of this uses one-time URLs that expire in minutes.
Whether third parties see the photo
Even if the app you signed up for keeps your photos private, they may be passed to third-party AI vendors for tagging or classification. Those vendors have their own retention rules, and their own logging, and their own risk profile.
The right answer is either "we don't pass photos to third parties" or a specific named list, with retention promises for each. Vague language is a red flag.
Why this is more sensitive than it sounds
Photos of your clothes may not feel private in isolation, but taken together they say a great deal. They reveal your body shape over time, your buying patterns, your income bracket, sometimes your address (mirror reflections, backgrounds), your habits, your relationships (someone else's clothes in the background). A well-motivated third party could learn a lot from a full wardrobe dataset. Most people would not consent to that if it were asked as a single up-front question.
A short list to check before you upload
- Does the privacy policy say photos are not used to train models?
- Does the app describe storage as "private" or "user-scoped"?
- Are third-party services named, with retention rules?
- Is there a clear way to delete all data, including backups, in a stated timeframe?
- Is there an option to export your data before you delete?
If more than one of these gets a vague answer, hold off.
What FitMatch does
Photos are uploaded straight into a bucket that belongs to your account. Only your signed-in device can generate a URL to fetch one, and each URL expires within minutes. The tagging model runs on our own infrastructure; nothing is passed to a third-party AI vendor. Your photos are never added to a training set. When you delete your account, images are gone from live systems within 30 days and from off-site backups within 90.
None of this is remarkable engineering. It is just the version of the app we wanted to use ourselves.