U.S. Copyright Office vs. Generative A.I.
Updated: Apr 13
Something positive and hopeful regarding the whole generative A.I. thing.
A friend shared this video with me last night. https://www.youtube.com/watch?v=y1LAvIJJP88
The info from the Copyright Office came out about three weeks, ago. What's stunning (and reassuring) to me is to hear them articulate from the legal perspective what many of us have been saying from the artist perspective (which "A.I. artists" still don't get).
Artists and the Copyright Office are both saying the same thing: When a machine produces a complex image based on a simple prompt from you, that is not you making art. The Copyright Office even literally uses the analogy of HIRING an artist, vs. BEING one.
Copyright Office: “When an AI technology receives solely a prompt from a human and produces complex written, visual or musical works in response, the “TRADITIONAL ELEMENTS OF AUTHORSHIP” ARE DETERMINED AND EXECUTED BY THE TECHNOLOGY—NOT THE HUMAN USER.” (my emphasis)
Copyright Office: “These prompts function more like instructions to a commissioned artist—they identify what the prompter wishes to have depicted, but THE MACHINE DETERMINES HOW THOSE INSTRUCTIONS ARE IMPLEMENTED IN ITS OUTPUT.” (my emphasis)
The loophole that is still good news (I feel): Though the direct output of the A.I. cannot have copyright protection (cannot “be copyrighted”), if you photobash individually un-copyrightable A.I.-generated images to create a new picture, that picture can be. This is the best one we can hope for, because this "loophole" is one that only skilled artists can exploit, thus limiting generative A.I. to actual TOOL status—if you don’t know how to make a good picture, you can't have A.I. do it for you. But if you do know how to make pictures, if you do have actual visual ideas, A.I. can alleviate a lot of the burden, just like other tools (digital and traditional, including photos), do for illustrators, concept artists, animators, etc.
But it’s even better than that: if someone uses generative A.I. to come up with a character design (e.g. evil robot, just picking something totally at random ), then separately does some painting(s) or drawing(s) of that character (say, for their comic), then even though the drawings and the comic are copyrightable, THE CHARACTER IS NOT. The character is in the public domain. In other words: not only can you not use generative A.I. to paint pictures for you, you can't design copyrightable characters (or vehicles, or environments) either. So, again, if you can’t design your own interesting stuff, generative AI isn’t going to bail you out. And “design” means CREATE VISUALLY, YOURSELF, not merely give a brief text instruction.
Companies that use A.I. are going to have to be very wary of the line between AI as a tool, and A.I. as an “unusable artist” (because all of that “artist’s” work is by definition in the public domain, and they NEED their visuals to have copyright protection). Someone may begin the concepting process for, say “evil robot” (just picking a concept randomly ) and then try to get away with copyrighting it by making only very small changes. But if someone else produces something similar or identical using A.I., in order for the company to defend its copyright it will have to prove that it didn’t in reality merely do the same thing.
If this policy holds I think it is REALLY great news for artists. Because the main concern about generative A.I. is that it is not a tool, but a replacement for artists. The copyright office is basically saying that is not possible. Because the simple math says if the image is predominantly coming from a machine then it is not protected.
Of course this doesn't mean things won't change, and that AI won't take over a lot of the workload. Companies and artists will figure out how to exploit it without sacrificing copyright. But if A.I. tools end up empowering vs. replacing artists, as new innovations have done in the past, the result can be growth (more content and more demand for it). It's important to note, though, that sometimes the increase in overall opportunities comes at the expense of specific previous types of opportunities, which sucks for some artists. I try to avoid the prediction game so that's as much as I'll say on that.
The video concludes by “the question of can A.I. be trained on legally copyrighted work is still unresolved, but they are looking into it.” So there’s that as well—but this outstanding legal question may prove to be less of an issue for the majority of artists in terms of the actual effect (still wrong, though).
What still surprises me is that so-called artists who fully rely on A.I. still don’t appreciate what this means in terms of their own creative involvement (or lack thereof). Everyone who knows anything about any of this is saying the same thing: YOU’RE NOT DOING IT. IT’S NOT YOUR CREATION. But the ego of the imaginary artist is not able to hear this.
I’ll repeat this anecdote: I once had a client say it didn’t make sense that a certain award for children’s book illustration go to the artist and not the client, because “I am the one who tells them what to draw.”
Now, whether the copyright extends to pig artists is another question entirely.