Return of the Robot Authors

Last summer I did a writeup of where copyright law stood regarding AI-generated images. You can find it here.

Now, to be clear, the parts of this issue that interest me are not the parts that have captured everyone else’s interest! Whereas the general discourse has focused on the peril to professional artists and on the use (or misuse) of copyrighted materials in training AI models, I have been focused on how the law treats the output. If you have an AI generated picture, who owns it?

Ownership of AI-Generated Works

After some hemming and hawing, I came to the conclusion that nobody gets to own the output! And the slowly grinding gears of justice have been bearing that prediction out. Copyright Office policy and Dr. Thaler’s court cases have shown that the machine itself is not regarded as an author, while a more recent Copyright Office decision has held that the person who prompts the AI to generate images for a comic book isn’t an author either.

I have a copy of the original decision at the bottom of this post, to satisfy those of you who hunger for knowledge. But before we get there, I want to point you to a couple of choice passages. The first is on page 9 of the decision:

Rather than a tool that Ms. Kashtanova controlled and guided to reach her desired image, Midjourney generates images in an unpredictable way. Accordingly, Midjourney users are not the “authors” for copyright purposes of the images the technology generates. … Because of the significant distance between what a user may direct Midjourney to create and the visual material Midjourney actually produces, Midjourney users lack sufficient control over generated images to be treated as the “master mind” behind them. (emphasis added, citations omitted).

In other words, the Copyright Office held that the AI is not just a tool used to achieve the human’s creative vision. It is distinct from Photoshop or a camera, in that the machine is generating images that the human did not actually conceive of. On page 10, the Copyright Office hits on the analogy to commissioning artwork from a human artist:

Instead, prompts function closer to suggestions than orders, similar to the situation of a client who hires an artist to create an image with general directions as to its contents. If Ms. Kashtanova had commissioned a visual artist to produce an image containing “a holographic elderly white woman named Raya,” where “[R]aya is having curly hair and she is inside a spaceship,” with directions that the image have a similar mood or style to a “Star Trek spaceship,” “a hologram,” an “octane render,” “unreal engine,” and be “cinematic” and “hyper detailed,” Ms. Kashtanova would not be the author of that image. Absent the legal requirements for the work to qualify as a work made for hire, the author would be the visual artist who received those instructions and determined how best to express them. (emphasis added, citations omitted).

One last point I want to emphasize, on page 11 the Copyright Office addressed the question of editing the AI-generated images, and concluded that the edits lacked the creativity necessary to establish authorship:

The changes to Zarya’s mouth, particularly her upper lip, are too minor and imperceptible to supply the necessary creativity for copyright protection. The Office will register works that contain otherwise unprotectable material that has been edited, modified, or otherwise revised by a human author, but only if the new work contains a “sufficient amount of original authorship” to itself qualify for copyright protection. Ms. Kashtanova’s changes to this image fall short of this standard. (emphasis added, citations omitted).

At the same time, the Copyright Office recognized that some parts of the comic are copyrightable. The text was written by Kashtanova and the images were arranged by her, so that the comic as a whole is protectable. The Office also suggested that edits to an AI-generated image could render the outcome protectable if they meet the standard for creativity. In the end, the Office concluded on page 12 that they would reissue the copyright certificate, but with an exclusion for “artwork generated by artificial intelligence.”

The Future

My kids have a variety of toys in the genre of “find the block that fits.” The law is a lot like that–we take the facts and figure out how they fit into the existing structure of statute and jurisprudence.

But as much fun as that game is, it’s kind of a diversion when you are looking at new technologies. We’re trying to fit the proverbial square block into a round hole. Right now I am pretty confident that nobody owns copyright for AI-generated images. But honestly, that’s a really weird outcome! It is what I, in my physics-soaked lexicon, would call an unstable equilibrium.

There is just so much money in the generation of media, from movies to songs to books, and I cannot believe that those monied interests will let the law stay in this state.

For example, imagine you are a huge media company that made its name in animation. It used to be that human hands had to slave over every cel to create an animated movie. Digital tools expanded what was possible, while also dramatically reducing the human labor required, reducing costs in a big way! AI could do the same thing, but the huge media company won’t touch it if there’s any uncertainty in their ownership of the output. If you were a huge media company, with a substantial and well-used lobbying budget, what would you do?

I fully expect that lobbying is already in the works to create some framework for copyright of machine-generated artwork. The lack of copyright protection is a glitch in the current state of the law, but it is one that Congress could easily remedy, for example by saying that the person who provided the prompt owns the copyright, that the person who owns the machine owns the copyright, or even by saying that the machine itself owns the copyright (looking at you, Thaler!).

Personally, I think it would be a nice future if nobody owned the copyright. We could easily make that part of the deal for the copyright monopoly–employ human artists if you want to actually own the art. So. Uh. Write to your congresspeople!

And for some light reading, here is the decision by the Copyright Office that I discussed above: