I understand your concern, but that's not how it works. At training time, the model learns condensed representations from the dataset. It doesn’t scan anything during image generation. In fact, it doesn’t really need to store anything other than the learned model weights.
ai ngambil artwork dari banyak artist (tanpa izin), terus dijadiin database mereka. abis tuh, ketika ada orang yg make ai pgn bikin gambar, misal "foto ala disney di pantai" algoritma ai bakal scan database hasil curian itu, terus semacam di "kanibal" buat jadi gambar yg baru.