A text-to-image model creates an image from a written prompt. A prompt naming a webcomic artist, for example, can produce a new illustration. That capability raises a question: when a model generates images associated with an artist’s work, what rights are involved? section 4.11
In the United States, copyright gives creators exclusive rights to their work, but fair use can permit some uses without permission. A pressing question is whether using artists’ copyrighted images as training data infringes those rights. The US Copyright Office says AI-generated images alone are not eligible for copyright protection; a work with AI-generated elements may qualify if it contains sufficient human creativity, assessed case by case. These are different questions: rights in training material and rights in the resulting image. section 4.11
Figure 4.7 An AI-generated illustration produced by an open source image model using the prompt “Sarah Andersen webcomic.”
An illustration generated from a prompt naming a webcomic artist makes the connection between image generation and artists’ rights concrete. section 4.11
