Federal judge says AI-generated artwork can't be copyrighted, because of monkeys

Originally published at: Federal judge says AI-generated artwork can't be copyrighted, because of monkeys | Boing Boing


This is assuming that “AI art” is 100% algorithm created. In practice it almost never is. To get decent results (like non-mutant hands) you generally need to manually modify it. The resulting work would be copyrightable, just like how those modified public domain works of literature (Sense and Sensibility and Sea Monsters, etc.) are.


Rather awkward for studios and networks that want to replace writers, actors, with AI, if they can’t copyright it.


… I for one welcome our new :monkey_face: :robot: overlords



In the absence of any human involvement in the creation of the work …

I’m sure greedy companies will try to exploit this statement. Does writing a mere prompt for AI satisfy this clause? What about modifying a single pixel of a generated image, a letter of a generated story, a bit of a generated sound, etc?


No, it isn’t assuming that at all. It is an uncontested fact of this specific case. From the article: “In 2018, he listed an AI system, the Creativity Machine, as the sole creator of an artwork called A Recent Entrance to Paradise, which was described as ‘autonomously created by a computer algorithm running on a machine.’”

So, two things here. First of all, Sense and Sensibility and Sea Monsters was not created using AI in any way, shape, or form to my knowledge, so I have no idea why you’re using that as an example. Second, if a human modified an AI created work, the resulting work might be copyrightable. As with most things in the law, it depends. In this case, it would depend on just how much human involvement there was. If the human involvement were minimal (minor touchups), I think it’s likely that the copyright office and the courts would find the work was not sufficiently the creation of human hands to warrant copyright protection. On the other hand, if the human made significant changes, then it probably would be copyrightable.


Another misleading headline. What the courts and copyright office have said over and over is that algorithms cannot be listed as the author of a work, a human must be listed as an author. But that doesn’t rule out a human using “AI” algorithms, as it always is. (The copyright office has also said that a human just giving a prompt may not be sufficiently “creative”.)


No, the headline is accurate. In this case, from the article, “A federal judge on Friday upheld a finding from the U.S. Copyright Office that a piece of art created by AI is not open to protection.” The issue was not that the work did not list a human author. The issue was that the work did not warrant protection because it was not created by a human. If the petitioner here had listed themselves as the author instead of the AI program, when they weren’t, that would also be problematic because it wouldn’t be true.


Given the whole purpose of copyright - to allow human beings to profit off their creative labor, the ruling makes sense. (“Prompt engineering” isn’t remotely at that level.)

The good news, this may cause some TV/film/music executives to wake up from their wetdreams of replacing creatives with neural networks and machine learning algorithms

That wasn’t really Hollywood’s ambition. The real risk was that executives would have “AI” fart out a “script” and then have a human “edit” it (i.e. write the damn thing more or less from scratch), but then not give the credit (or the money) to the actual, human writer. The idea that the “AI” could write anything usable would just act as a pretense to deny writers proper compensation for their labor. This also likely causes those executives to wake up from that particular dream, too. (Although the outcome of the current strike will probably do that job all on its own, for Hollywood at least.)


I’m surprised that I need to spell this out, but if you really aren’t following, here goes: “Sense and Sensibility and Sea Monsters” is copyrightable even though it is based on a public domain Jane Austen novel because it contains new material unquestionably copyrightable because it is generated in the current day by a human. Similarly, if you modify an AI-generated drawing, even assuming that the ruling stands that such drawings are public domain, the resulting drawing is clearly copyrightable because any human modified public domain work is copyrightable. The issue of AI is irrelevant other than in the suggestion that an AI drawing is, like Jane Austen in the 21st century, public domain.


I would wager that the majority of people generating AI art today aren’t making substantial, transformative changes to the imagery the AI spits out though. Some subset of professional artists, designers and illustrators do, sure. But a helluva lot of people don’t.


The fact that Adobe is incorporating generative AI into Photoshop implies that at least they see modification of AI-generated art as the future of illustration.

That would be the subset of professional artists, designers and illustrators I just mentioned.

This court ruling doesn’t really seem to be about those people.


As Auguste Comte sarcastically (although accurately) noted, the law applies to everyone and both the rich and poor are forbidden from sleeping under bridges. Similarly, this ruling (although clearly based on an submission from a tech company) doesn’t make any exceptions for professional artists using AI-based tools as opposed to the much derided “techbros”.

1 Like

Aside from outfits like the one that brought this case (who I can only assume were looking to prove something; given the vanishingly small odds of the copyright office raising any doubts if they’d just said it was a work for hire by one of their employees or contractors rather than openly adding an element of novelty to their claim); the biggest potential impact seems like it will be on the sellers of ‘AI’ services and middleware; (whose output typically isn’t touched by a human before going to the customer) rather than sellers of final media products looking to trim their artist budget.

If bot product could be copyrighted as though it were a work for hire that would put OpenAI/Midjourney/etc. on a substantially stronger footing in terms of how they sell their services since that would allow them to either retain the copyright and grant the customer a more or less limited license for certain uses; or, for customers paying enough, make the customer the assignee for works generated for them.

Since that is not possible they have a lot less to work with: they can still burn your account for ToS violations; and it’s not as though contract law is completely toothless; but that’s a significant step down from having the copyright of the output being something you can either offer as an inducement to pay for a relatively expensive full-commercial-use tier or retain to provide a legal foundation for placing restrictions on lower usage tiers; since having your permission be required to keep derived works from being unauthorized is pretty punchy.

Given the significant cost of operating some of the larger and more sophisticated models there seems to be a lot of interest in SaaS-ing them; so this seems like a pretty substantive decision(though an expected one; the ‘monkey selfie’ case was a bit of a media circus but the creativity standard that led to it was old, old, news that just didn’t come up very much) for the interface between the SaaS vendors and their customers; since that is one where a lot of transactions are occurring and the structure of the agreements could have looked a great deal different if the product being sold were a copywriteable work-for-hire rather than just a compute service with a ToS.


I suspect that, in practice, there could end up being a nontrivial difference in handling because the Copyright Office, at least historically, has treated claims of human authorship as presumptively valid more or less across the board. Unless someone else decides to litigate and open the question(something along the lines of Feist Publications, Inc. v. Rural Telephone Service Company, Inc.), or you walk right into it like the monkey selfie guy by publicly admitting that the monkey did it and that it wasn’t part of your creative plan to use local simians as a medium of artistic expression; they aren’t in the business of hassling you over whether you were feeling super creative at the moment of fixation.

That’s going to mean that people(including ones whose claims almost certainly shouldn’t pass muster) who make copyright claims in the traditional style, with a human in the loop and so on, will go largely unscrutinized, unless someone decides to push the issue after the fact; while any attempt by people like the guy in this case; or by an outfit that insinuates that there’s a human in the loop when talking to the copyright office; but mysteriously offers API turnaround within a few thousand milliseconds, is going to be judged rather differently.


True, although much like with ChatGPT, people seem to miss that the future doesn’t really seem to be companies providing this as a service given that both generative AI like InvokeAI and LLMs like llama.cpp can run on very modest local hardware. Both for privacy and cost reasons the future of AI is likely to be local, not in the cloud, and quite likely open source at that.


Though I agree with most of what you’ve said here, I have to take issue with this. It’ll be proprietary, expensive as hell, and companies will guard and control it jealousy, which means many will absolutely strive to keep it “cloud” only. There will be very little open source left when all is said and done. Ever has it been that a new idea comes along, cheap and often open source, and within a few years is gobbled up by the big companies and they slowly raise the price until everyone is cooking. See: The Cloud. See: Streaming channels.

AP (algorithmic processing, because it’s still not artificial intelligence) is in its wild west phase right now. But you even pointed out Adobe is including this now as an offering, which very much indicates the direction it’s going to go. No giant corporation is going to let a profitable horse run wild for long.


Unfortunately for us, nobody copyrighted “Condescending Opening Responses.”