Your child's photo in an AI book: what happens? | MyOwnChildbook
You’re about to upload a photo of your child to a service that turns it into something with AI. Before you click “upload”, there’s one question most parents never ask out loud: where does that photo actually go? It’s a fair question, not just for a children’s book, but for any AI tool that asks for a photo of your child.
What actually happens to an uploaded photo
Most photo-based AI services fall into two categories, and the difference is bigger than most people realise. Free consumer apps on your phone often partly fund themselves by using user data, including uploaded photos, to further train their models, unless you actively opt out in the settings. Paid services running through a business API connection usually sit under a different regime.
OpenAI, the maker of the gpt-image-2 model that many AI children’s-book services rely on, is explicit about this in its own API data documentation: data sent through the API is not used to train OpenAI’s models, unless a business explicitly opts in to share it. By default, abuse-monitoring logs are kept for up to 30 days, unless a business has agreed to stricter zero-data-retention terms.
Why the difference between a consumer app and an API matters so much
That distinction isn’t a minor technical footnote. European privacy law (GDPR) requires companies to practise data minimisation: keep only what’s needed, for no longer than needed (GDPR Article 5). A company that routinely kept children’s photos without a concrete reason would be breaching those principles. Parents also have the right to demand deletion of their data, the so-called “right to be forgotten” (GDPR Article 17).
In practice, this means a carefully built AI pipeline treats a photo as a temporary tool, not an asset worth collecting.
How our own pipeline actually handles it
Edwin, the data engineer who built MyOwnChildbook’s illustration pipeline, explains what that looks like in practice: “The photo you upload gets sent along as a reference image in the call to gpt-image-2, so the model can approximate hair type, hair colour and skin tone. Once the illustrations are generated and approved, the photo is automatically deleted from our servers. We never use uploaded photos to train models, ours or OpenAI’s.”
That’s exactly why the split between “free consumer app” and “business API integration” matters in practice: as a father of two, Edwin wanted that guarantee himself before he built it.
When this doesn’t feel right for you
Some parents won’t find any guarantee reassuring enough: the idea alone of sending a photo of their child into a cloud somewhere doesn’t sit well. That’s a completely legitimate boundary. Policy and engineering can reduce risk, but they can’t erase the fact that a photo briefly enters a system you can’t see inside yourself.
For those parents, a book built from a description rather than a photo is a more honest fit. Wonderbly, for instance, traditionally personalises books through a chosen name, hair colour and skin tone, without any photo upload involved. Our own comparison of Wonderbly and MyOwnChildbook lays that difference out side by side.
Not blind trust, but traceable choices
Trust in an AI service shouldn’t hinge on a nice-sounding promise on a website. It should hinge on traceable facts: which model, which API terms, which retention period, which right to deletion. Our explanation of how gpt-image-2 creates an illustration shows exactly what happens to that photo once it enters the pipeline.
That’s ultimately why the question “where does this photo go?” is one you’re always allowed to ask, of us and of any other service that asks for it.
