ThePaintingOwl wrote:Maybe in your specific case you didn't do a standard work for hire contract but typically that is part of any serious commission work, such as D&D book art, and copyright goes to the company that paid for the work not the artist.
There's a reason artists are credited in
DnD books, and even Magic Cards give the artist's name. There's a reason you know their names at all (while you probably know no names of any given chemist who hasn't won a Nobel prize). Wizards buys rights, not copyright, to the individual image. This is true broadly of the entire publishing/illustration industry. This is the case of (2) in your link: "when a certain type of work is created as a result of an
express written agreement between the creator and a party, specially ordering or commissioning the work." Note the bolded section and ask yourself why it's there and who probably owns the copyright lacking an express written agreement.
Or you could just read their clarification at the end of page one;
The work made for hire concept can be complicated and has serious consequences for both the individual who creates a work and the hiring party who is considered to be the author and copyright owner of that work. This circular draws on the Copyright Act and judicial interpretation to provide a general introduction to this topic and answer common questions.
Because there's a lot of law around work for hire and copyright, and it gets fucky because, as an example, I can own the copyright of the picture book
The Amazing Mr. Mazy and the Fluffy Hat which has images in it, while at the same time, the illustrator owns the copyright to the images themselves (which is the norm for most picture books illustrated by one person but written by another). This is how you can, not so hypothetically, own the copyright to a comic book but
not the copyright to the artwork within said comic book.
Most commission agreements expressly keep the copyright to the artists and only give commercial or fair use licenses to the commissioner. They expressly do not automatically infer copyright to the commissioner automatically. It's defined by agreement. Disney owns the copyright to the Avenger's movie, but you can bet that when Hans Zimmer makes a musical score for a movie, he still owns the copyright to the music which is why his name is on the albums credited as the artist (along with whatever orchestra did the playing).
I think you might be confusing it with other fields and industries, where this is often the case. Pharma for example, generally contracts the work to individual researchers, but explicitly owns their work product. Publishing can do the same thing, but market forces have generally resulted in artists having more leverage over keeping copyright than scientists in research fields hanging onto patents. That would define (1) in your link.
Consider Midjourney's
TOS in this regard, which (aside from extreme weasel wording on 'assets' and 'ownership' that I'd point to and warn anyone using it about what they think their rights are) expressly tells you to consult a lawyer about the laws of your jurisdiction, and nowhere in the
TOS is it expressly stated that you own any copyrights. That's a change from their original
TOS, which did expressly state that, but they cut the difference as case law became complicated.
This is not me making a hypothetical argument by the way. That's literally what courts are saying, and their understanding of AI is actually surprisingly sophisticated in how they examined in use in the two cases I know of. The comic book case hinged on a very sophisticated ruling examing 'human effort' as it's defined in precedent.
The only reason AI is being treated differently is people anthropomorphizing it as "someone else doing the work for you" instead of a single human using a software tool to generate output.
You keep saying that and it still has nothing to do with what I'm talking about.
You don't have to anthropomorphize AI to accuse users on one end of not doing any work. Pressing
ctrl+a,
ctrl+c, and
ctrl+v is not a creative process, which isn't the only way generative AI can be used, but it's the way some people have tried to use it only to find they have no copyrights since they didn't make anything. It could be sidestepped, hilariously, by putting in even a modicum of extra effort because of the way the law is written, though.
US Copyright laws actually cover that and there's case law precedent because image generation is a lot older than generative AI. This stuff goes back to the 90s, and even further back because of people like Andy Warhol and James Patterson and Tom Clancy leading the way on the question of 'who owns what and why do they own it?'
"Market impact" or "quality of work" are irrelevant in determining if a particular work is protected by copyright.
See the previously cited fair use/copyright case I linked. The SCOTUS decision hinged a lot on questions of why IP rights exist and what their express purpose is. Marketability and commercial use, as well as quality examinations, were all central to the case.
Legal interpretations are almost entirely about
why a law exists past a certain point.
This is what I've been saying about the current approach being a dead end.
Consider that the current iteration is probably a dead end that has already reached its real limits in what it can practically do, but it won't end there. The models in use now will be used to develop more complex and sophisticated models later. Indeed, the current implementation is a necessary stepping stone to more advanced systems. Getting a computer to spit out coherent sentences and responses, even if they're often factually incorrect and start falling apart after about 750 words, is a necessary starting point for doing more than that.
I think people are overestimating what it can do now and how fast it'll develop (electric cars are 150 years old, they're still not market standard and are only just now becoming very common in the past decade, as an example) but the systems will keep developing so long as there's interest or money to be made.
Stuff like Stable Diffusion and Midjourney are probably dead ends, but integration like what Adobe is doing (and notably, Adobe is sidestepping basically all the obvious legal pitfalls around copyright and fair use) will become more sophisticated. Generative AI in a pure form is a dead end, but mixed and multimedia AI uses are going to keep advancing.
