Another way to think about how AI image generation works is to know how art students train.  Go to any art museum and often you will find people sitting in front of a piece by Picasso or Van Gogh, and they are drawing it.  No, they are not doing so to plagiarize the artist.  Instead, they are drawing it to study it, to understand how Van Gogh used the colors he did or how Picasso drew the outline of a woman’s hip.  That does not mean they will go one to create art “just like” the artist they studied.  Instead, they will develop their own style which will be influenced in various weights by the artists they studied. 

An AI image generator model is that art student.  As the person instructing it, I am using prompts and settings to influence how it will create an image and with that and the data it has collected, it will create what it thinks I want.  Sometimes it is spot on; more times, however, it requires more from me.  In fact, the challenge to me is to have what I want well pre-visualized in my head before expending computing units and giving the AI model its instructions.  And yes, sometimes the model will return an image that looks as if it has an artist’s signature in the corner – no, it did not copy an artist's work, it is just that it has seen a picture with something like that so often, it thinks it has to be part of what it makes, in essence “signing” its own work!