Dear Doc:
I see that several artists and authors have sued artificial intelligence companies, claiming that using their writings, artwork, voices, etc. to train AI systems violates copyright. What do you think of that? Can they stop this whole AI thing using copyright law?
Signed,
Asking for an Artificial Friend
Dear Asking:
The Doc is skeptical of these lawsuits (See articles in New York Times and Art News) for many reasons. In order to understand why, let’s look at what copyright is and why we have copyright laws.
Copyright (at least in the United States) is a federal law established under Art. 1, Sec. 8, Cl. 8 of the Constitution which says that we have copyright to “promote the progress of science.” As you know if you’ve been keeping up in class, “science” in this sense is how Enlightenment thinkers viewed it. That is, it means general knowledge, and not the physical sciences, which Mr. Jefferson et al. called “natural philosophy.” In order to achieve this progress, the Constitution says that we should secure to authors “exclusive rights” for “limited times.”
Although originally copyright only applied to written materials produced and sold in the United States, and lasted just 14 years, Congress has repeatedly expanded the scope and duration of copyright by covering music, photographs, motion pictures, semiconductor masks, boat molds, and other types of works (17 U.S.C. §102), as well as making it last for the life of the author plus 70 years, and prohibiting more than just making copies (so that it now prevents making translations, adaptations, performances, digital transmissions, and the like.) (17 U.S.C. §106)
One thing that has never been prohibited under copyright law is learning from a protected work. The Doc thinks that is obvious, since if you want to promote the progress of knowledge, then studying and learning from earlier works is a great way to do that. So when we use works to educate (or “train”) an artificial intelligence system, the Doc thinks that’s just what we are supposed to do under the system set up by the drafters of our Constitution.
“But,” the Doc hears you sputtering, “isn’t taking someone’s work without their permission and using it in your AI just taking the value of what they have created? Isn’t that just theft?”
The Doc wants you to think about that. If I publish a book on the Internet, and you read it, and it changes the way you think about something, what have you taken from me? Ideas? Well, ideas are never protected by copyright. 17 U.S.C. §102(b) says, “In no case does copyright protection for an original work of authorship extend to any idea, procedure, process, system, method of operation, concept, principle, or discovery, regardless of the form in which it is described, explained, illustrated, or embodied in such work.”
What if you steal my actual, real book, read it, and then put it back on my shelf? Well then, you’re a thief, to be sure, and I may be able to recover the value of not having the book for the time you took it, but again, the ideas are free. It’s just the physical object that I am compensated for not having for some period.
“But aren’t ideas valuable?”, you may ask… “And what if the AI takes my unique style, huh?”
Yes, says the Doc, but in our system, just because something has economic value or is unique and recognizable does not mean that there is an automatic legal right created. This “if value, then right” regime could create lots of confusion since nobody would have a clear idea of what was permitted and what was prohibited under law. Rather, in our system, legal rights are defined by statute and common law, and value often flows from the legal rights (but sometimes, not).
Those folks who are bringing lawsuits against AI companies for using their copyrighted materials are going to have to show which specific rights provided under the law have been infringed (for instance, is the stuff made by the AI substantially similar to something that they have registered with the Copyright Office), and also will have to argue that any use of their materials is not “fair use” (17 U.S.C. §107).
In a recent hearing, Judge Orick of the U.S. District Court for the Northern District of California voiced his concerns. “I don’t think the claim regarding output images is plausible at the moment, because there’s no substantial similarity [between the images by the artists and images created by the AI image generators].” According to a motion filed by the AI company defendants, “Plaintiffs’ direct copyright infringement claim based on output images fails for the independent reason that Plaintiffs do not allege a single act of direct infringement, let alone any output that is substantially similar to Plaintiffs’ artwork…Meanwhile, Plaintiffs’ allegations with respect to [Artist] Andersen are limited to only 16 registered collections but even then, Plaintiffs do not identify which “Works” from Andersen’s collections Defendants allegedly infringed.”
The issue is that the new AI models like Stability appear to copy artists’ styles, but “style” is not protected under existing copyright laws, leaving a kind of loophole that AI image-generators can exploit. Artists are trying to use Copyright law to protect style, in ways similar to the lawsuits brought by music artists (or their estates), claiming that just because one tune is recognizably similar to another, there is copyright infringement going on. The Doc thinks that this is a long shot, but like any lawsuit, it will take a while, and the lawyers will get rich regardless of who prevails. Stay tuned, and the Doc will keep you up to date as things develop.
Do you have some intellectual property that’s being used by someone, somewhere, for some purpose with which you disagree? Give the attorneys at LW&H a shout. They enjoy defending our copyright system, and the Doc hears that they’re pretty good at it.
Until next month,
The “Doc”
— Lawrence A. Husick, Esq.