Do creative works that utilize generative AI require attribution?

Question

Suppose I create content (videos, essays, etc.) that utilizes generative AI to create illustrations (e.g. Midjourney, DALL-E, etc.), do I need to cite the AI tool I used? Will I still be the copyright owner of the overall work (not including the generated images)?

Answer 1

Do creative works that utilize generative AI require attribution?

Attribution is a plagiarism concept relevant to academic ethics and in certain other non-commercial settings like contests. Plagiarism is not copying without permission (a copyright concept), it is copying without attribution. Unlike copyright law, the plagiarism requirement of attribution applies to copying ideas without attribution, as well as copying final expressions of ideas (which when done without permission violates copyright law).

Attribution is sometimes required by union-labor agreements in industries like the movie, television, and music industries, and can be necessary to assign "moral rights" in countries that have them as part of their copyright laws.

In circumstances where attribution is required for one of these reasons, attribution of an AI source for the material is also required.

Many U.S. Courts are also seriously considering imposing a rule to require a statement of whether AI was employed in drafting legal documents that are filed with them.

Will I still be the copyright owner of the overall work (not including the generated images)?

Generally yes. See also this prior Law.SE answer, which is linked to in the comments.

Answer 2

If you want to register your work, you need to declare usage of AI

In the somewhat recent decision of the copyright office in re: Théâtre D’opéra Spatial, an artist was denied copyright registration for not declaring that he used an AI to generate a part of the image and tried to not disclose that. The board found:

After carefully examining the Work and considering the arguments made in the First and Second Requests, the Board finds that the Work contains more than a de minimis amount of AI-generated content, which must be disclaimed in an application for registration. Because Mr. Allen has refused to disclaim the material produced by AI, the Work cannot be registered as submitted.

Likewise, earlier, the copyright office revoked a registration earlier with in re: Zarya of the Dawn for usage of undeclared AI content, and re-issuing a registration only for the non AI-generated text:

We conclude that Ms. Kashtanova is the author of the Work’s text as well as the selection, coordination, and arrangement of the Work’s written and visual elements. That authorship is protected by copyright. However, as discussed below, the images in the Work that were generated by the Midjourney technology are not the product of human authorship. Because the current registration for the Work does not disclaim its Midjourney-generated content, we intend to cancel the original certificate issued to Ms. Kashtanova and issue a new one covering only the expressive material that she created.

Declaration of AI-generated content is not to be confused with attribution. It is a declaration that you did not create this aspect and do not claim copyright in that aspect. Attribution is academically telling who made the material you cite.

You can not own copyright in AI-generated content in the first place, but you can have copyright in any other aspects of a work, such as the text in the case of Zarya of the Dawn.

You can't own the AI parts

As both cases above show, you can't own copyright in the AI made parts.

You do own copyright in non-AI parts

As the Zarya of the Dawn case shows, as long as your own work is not de minimus and shows the required creativity, you do own the copyright in those parts only. In other words, if you use AI-generated images in a collage, you own the arrangement of them. If you do a comic, you own text and arrangement, and so on.

If your work contains a few small (de minimus) AI-generated parts, but is mostly your own work, you do have full copyright, following the Théâtre D'Opéra Spatial decision. A possible case would be if you used AI to generate just a pose (akin to a wood puppet) or rough sketch of an arrangement and then did the actual painting that covered it completely with expression you created, that would probably count as de minimus. Similarly, you might apply an AI-driven filter over a finished work to get a specific look or effect without lessening your copyright.