How to Attribute an AI Image
Who is the author of an AI Image? One of the moral rights associated with copyright is the acknowledgment of the work's creator. However, when an AI model is the creator, authorship rights and credits are not clear, especially when the training set comes from many artists' work.
In this article, we describe how to attribute an image generated by an AI model as a guide for editorial teams, students, and other digital writers.
How to Attribute an Image Generated on Tess
As opposed to other AI models, Tess images are fine tuned on a single person's work, trained in partnership with an artist in a visual style that they have crafted. We believe that the original artist owns the copyright to the image, which they transfer to the customer through a licensing agreement.
When using an image downloaded from Tess, we recommend attribution to the name of the Tess model. Linking to the Tess model gives the artist exposure and the opportunity to generate more passive income through their Tess model.
The artist has control over the name of their model. Some artists choose to give it their real name, and others choose a generic name or pseudonym. Tess honors the artist's name and reputation by giving them agency on how their visual style is attributed.
How to Attribute an Image Generated by Dall-E, Firefly, or Midjourney
Most copyright advocates would say that AI generated art work cannot be copyrighted. However, given that there is human input involved in AI generation, this feels strange. At the time of writing, different sources, like the Brown University Library and RMIT, have different suggestions of how to attribute AI-generated work as this is an evolving area of law and copyright understanding.
Here are few approaches for AI-generated image attribution:
- Attribute the model and reference the prompt: "Image generated using Adobe Firefly from the prompt 'interstellar purple and orange space squid.'"
- Attribute the prompter and tool used: Photo by Jane Smith/ Adobe Firefly
- A more detailed attribution of the model, version, and prompt: "Description of chat" prompt. Name of AI tool, version of AI tool, Company, Date of chat, URL.
We believe that approach 2 (attribute the name of the prompter and the tool used to generate) is the cleanest and most clear model. This approach accommodates the changes that the prompter might apply manually or technologically after generation.
Nonetheless, ongoing court cases show that choosing a properly-licensed AI Generation tool like Tess is a safer, more ethical approach for publishers. Prompters using Midjourney, Dall-E, or similar tools must be conscious to avoid violating copyright or reproducing copyrighted elements.
I hope this article clarifies how to attribute images generated using AI – happy prompting!