Stephen Thaler’s Legendary AI Copyright Losing Streak Ends With Nowhere Left To Appeal
from the public-domain-ftw dept
We’ve been covering Stephen Thaler’s quixotic quest to get copyright (and patent) protection for works generated entirely by his AI system “DABUS” for years now. If there’s one thing Thaler has proved beyond all reasonable doubt, it’s that you can be comprehensively, thoroughly, and repeatedly wrong at every level of the American legal system and still keep going. He loses everywhere, every time, at every level. The Copyright Office rejected him. A federal district court rejected him. The DC Circuit rejected him. The Patent Office rejected him. Courts rejected his parallel patent claims. Even the Trump administration—not exactly known for its nuanced intellectual property positions—told the Supreme Court not to bother hearing his appeal.
And now, the Supreme Court has declined to take up the case, putting the final period on what has been one of the most impressive losing streaks in recent IP law history.
Plaintiff Stephen Thaler had appealed to the justices after lower courts upheld a U.S. Copyright Office decision that the AI-crafted visual art at issue in the case was ineligible for copyright protection because it did not have a human creator.
That was always the fatal flaw with his argument. He wasn’t making the more nuanced claim that a human who uses AI as a tool should get some copyright protection. He was making the maximalist claim: the AI did it all by itself, and it (or rather, he, as the AI’s owner) should get the copyright anyway.
The image in question—”A Recent Entrance to Paradise,” of train tracks entering a portal surrounded by green and purple plant-like imagery—was, according to Thaler, created entirely by DABUS with no human creative input. Every single institution that looked at this said no.
A federal judge in Washington upheld the office’s decision in Thaler’s case in 2023, writing that human authorship is a “bedrock requirement of copyright.” The U.S. Court of Appeals for the District of Columbia Circuit affirmed the ruling in 2025.
Thaler’s lawyers, for their part, tried to argue that the stakes were too high for the Court to sit this one out:
With a refusal by the court to hear the appeal, Thaler’s lawyers said, “even if it later overturns the Copyright Office’s test in another case, it will be too late. The Copyright Office will have irreversibly and negatively impacted AI development and use in the creative industry during critically important years.”
That’s rich. The Copyright Office is already working through the genuinely harder questions in cases involving tools like Midjourney—cases where humans actually did have meaningful creative input. Those cases are moving through the system right now. The problem for Thaler is that he chose the worst possible vehicle to force a Supreme Court showdown: a case so maximalist in its claims (the AI did everything, humans did nothing, give us the copyright anyway) that courts could rule against him on the narrowest possible grounds without ever having to engage with the nuanced questions at all. His all-or-nothing bet made this an easy case.
Still, the question of what happens when a human uses AI as a creative tool—rather than letting the machine do everything—isn’t actually as novel or unsettled as many people seem to think.
Copyright law has required human creative choices since at least Burrow-Giles Lithographic Co. v. Sarony all the way back in 1884 in a case about whether or not photographs get covered by copyright. And the wonderful Feist Publications v. Rural Telephone Service from 1991 (a case we cite often) hammered the point home by establishing that copyright demands original creative expression. Consider how this already works with photography. A photographer who frames a shot of a landscape gets copyright protection in the creative choices they made—the composition, the angle, the timing, the lighting. But the landscape itself? No human created that. It gets no copyright. The camera mechanically captured what was in front of it, but the human’s original creative decisions (and only those original creative decisions) are what copyright protects.
AI-generated works should work roughly the same way. If a human’s creative input—through a sufficiently specific and expressive prompt, through selection and arrangement, through iterative creative choices—meaningfully shapes the output, that human contribution can be protected. But the parts that the AI generated autonomously, without human creative direction? Those are “the landscape.” They’re the thing no human authored.
There will certainly be disputes at the margins about exactly how much human input is enough, and where the line sits between “I told the AI to make something cool” and genuine creative direction. But the fundamental framework for handling this already exists. We’ve been here before with every new creative tool, from cameras to Photoshop. The principle has always been the same: copyright protects human creativity, regardless of the tool used to express it.
Thaler chose to fight for the one position that had no support in law (or in common sense). His losing streak is now complete, and there’s nowhere left to appeal. But the legacy of his many, many losses is actually kind of useful: he has, through sheer persistence, generated an incredibly clear and consistent body of authority establishing that purely AI-generated works, with no human creative input, do not get copyright protection.
So, thanks for that, I guess. Oh, and I guess we can confidently post that “Recent Entrance to Paradise” image as it, like the monkey selfie before it, is officially in the public domain.
Filed Under: ai, copyright, copyright office, copyrightable subject matter, dabus, human creativity, stephen thaler, supreme court
Comments on “Stephen Thaler’s Legendary AI Copyright Losing Streak Ends With Nowhere Left To Appeal”
The problem with protecting the human contribution to an image generated by “AI” is the inability or difficulty to isolate the human-authored parts in the final product. That might only cover the text of the prompt, and even then, the text may not be unique to warrant copyright protection. If you prompt an image generator using the phrase “old man,” there’s no way anything useful is copyrightable there. If you used a human-authored image in img2img processing only the original image would be protected and only for the original copyright holder who would already hold it, so the new product wouldn’t really qualify either, even if it integrated human-authored work.
This is easier in text where human-authored phrases can be copied from the work and isolated, similar to how an older work in the public domain can be remixed and only the novel parts copyrightable, such that Sherlock Holmes as a character can’t be subject to copyright anymore, but new characters invented by a new author could be.
Honestly, one of the real benefits of the nature of “AI” “art” is the lack of copyrightability. It means companies that use it can’t copyright those uses. It gives them more incentive to use human artists and writers if they want the traditional control they like having for monetization and licensing, but also makes it harder for them to suppress fan art and fiction and remixes if they do use “AI.” It becomes a devil pact with terms they may not like or at least, a means by which the public gets more access to culture earlier with less threat of lawsuits, even if the work may be less valued due to its origin.
That said, I’d prefer the art be human-authored. Humans know how to make art already. We didn’t need robots to do that. We need to use the LLMs to do the things beyond human capabilities like reading all the research on a type of cancer to spot patterns for future possible treatments, and then verify their work thoroughly before trusting it.
Re:
But that applies to other things, and American law is not consistent about it. If a human gives a camera to a monkey, there’s no copyright, even if the human had significant contribution—such as by setting up a scene, lighting, etc., for the monkey to photograph).
On the other hand, we let television companies copyright videos of sports broadcasts. The camera operator follows the ball or puck, as every sports camera operator does, and somehow that’s enough. It may well be the only reason those people haven’t been replaced by computer-controlled camera motors; it wouldn’t be hard to embed a motion-tracker in the object that’s the focus of the action.
And every tourist taking the same clichéd tourist photograph (like posing with their arms out to imitate a missing “T” in a city-name sign) gets their own copyright, for some reason. Why not the tourist bureau who staged everything to get exactly that photograph?
Sometimes, we kind of don’t know how to make art, in that the novelty of someone figuring out a new thing to do is what makes it art—like Duchamp’s “Fountain”, Thad’s recent comment on the article about teaching students to write worse, or Max’s “brilliant existentialist essay” (a blank page) in the 2008 Get Smart film. Maybe something generated by a computer will eventually be viewed the same way.
Re: Re:
This is less of an issue of copyright and more of an issue of access and monopoly. If everyone could film a sports event, the corporate recording would be less valuable, but they get a monopoly on access to film the event, so it’s artificially scarce. The exclusive access, not the copyrightability of the recording made by human camera operators, is the issue.
By that argument, why wouldn’t a person get a copyright on their own appearance if they showered and dressed themselves that day or even more, applied make up or styled their hair or dyed it?
The novelty of Fountain wasn’t a new method of creating art, but rather a bold method of arguing and interpreting what is art. And humans did do that without help from a machine.
Thad’s recent comment was human commentary on current events, something humans have done for millennia. Read old Greek and Roman writers and you’ll find contemporary comments about in-jokes and cultural topics.
Note my emphasis in the quote. There’s a difference between the source/method of creation and the perception of the work. Viewing LLM-generated work as “art” in a particular way (i.e. assigning it meaning) is the act of an observer or critic, which doesn’t necessarily require knowledge of the method of creation. Humans experience apophenia, even intentionally, all the time. Sometimes even misunderstandings are the source of new creative ideas. And again, that’s in the perception, not the creation.
Re: Re: Re:
You’re making a good argument for why it should not be copyrightable. Not only is there little creativity, there’s no public benefit to the extra monopoly of copyright on top of their general monopoly.
So I think copyright definitely is an issue, if not the issue. We need to rethink the whole thing. Who’s hurt when I give a friend a copy of a 1993 basketball game? This is Thaler’s failing too: treating copyright as just a natural natural to be slightly extended, and ignoring the stated purpose of the law (to benefit the public, not the artists or whatever we want to call Thaler).
We should be making the Librarian of Congress justify every area in which copyright is to apply, every three years, with public hearings and everything; with anything not properly justified in that period being exempt from copyright.
Yes, exactly. The legal standard doesn’t make much sense. If you arrange a bowl of fruit and take a picture of it, you get a copyright. Why? Because the law considers the styling, “make-up”, and such to be creative choices of the photographer, even if they didn’t really make any.
I have to believe that was probably the very point. The point wasn’t to win, it was to set certain decisions in stone by losing in a particular way.
Hopefully that fallen sword proves useful when Disney tries to argue that AI which is merely capable of generating works with
Mickey MouseDonald Duck content on demand is still covered by its copyrights.Re:
You know what they say: the most convoluted explanation is probably the correct one.
Trainwrecks happen
The AI shoud have seen it coming.
Code Copyright
Does this mean that companies can no longer copyright the code that AI writes? So anyone could copy their software?
Re:
No; it means the generated stuff never was copyrighted. And, sure, anyone with access can copy it.
It’s imprecise to say that companies copyright stuff. They don’t; the government grants them a copyright, maybe.
Re:
Not generally, but Microsoft is indemnifying users of their own tools from copyright infringement to some degree. If you can get it to spit out bits of the Windows source code without trying to get it to infringe, then I guess it’d all be nice and legal.
Will AI coding agents lead to public domain software?
Meanwhile, I read a report on Ars Technica – https://arstechnica.com/ai/2026/03/after-outages-amazon-to-make-senior-engineers-sign-off-on-ai-assisted-changes/ – which, in the context of this article and comments, makes me wonder about a few things.
Assume, for the moment, that Thaler’s Quest becomes precedent meaning that, absent “sufficient” human contribution, there’s a legal presumption that any product of any AI agent lacks any and all IP protections.
Do “senior engineers” signing off on “Gen-AI assisted changes” add “sufficient” human contribution for the code changes to acquire IP protections, or is their work merely akin to a human looking at “Recent Entrance to Paradise” and saying, “I {love|hate} that image” (or “if that’s paradise, I think I’ll wait for the next train”)?
Over time, as more and more contributions to a corporation’s code base originate with AI agents, at what point does the entire code base lose IP protections and become public domain? Is it triggered on the first AI contribution or, at the other extreme, will a single remaining human-written module be sufficient to maintain a corporation’s IP claims to the entirety of the code base? Or is it somewhere in between?
If private ownership vs public domain depends on the proportion of lines of code that had human or AI authorship, that’s going to be a tricky (but not impossible) thing to track. It would also likely mean that accountants will want to revalue IP assets downwards to accurately reflect the proportion that the corporation actually owns. Will we see shareholder lawsuits based on corporations mischaracterising the true origins of lines of code in their software products?
Maybe Mr Wilson is right. Maybe companies will have more incentive to use human software engineers if they want the traditional control they like having for monetization and licensing – or maybe society will become the winners when things like Office and Creative Suite suddenly fall into the public domain.
Re:
95 years after publication, or 120 years after creation, whichever comes first. No amount of added public-domain code will invalidate the original copyright(s).
It would have been amusing if DABUS not Thaler was granted the copyright, which then becomes a moot point, when Thaler is be unable to show that DABUS assigned any of the rights to him, he could even be unable to use the work, with added fun, if he ever turned off or reprogramed DABUS, which would constitute the end of the artists “life”, thus rendering the piece an orphan work, whith the copyright clock ticking.
Re:
Oh it would be even better than that, after all if DABUS was ruled to be enough of a ‘person’ under the law to qualify for copyright ownership then it certainly seem that a case could be made that attempting to reprogram or shut them down would constitute attempted murder, and that’s not even taking into account other legal implications such a ruling might have…
‘After careful consideration we have concluded that DABUS qualifies as a ‘person’ under the laws applying to copyrights, and thereby assign the copyright to them as of this point. Now about that legally binding contract which I’m sure you already have where they assigned all their copyrights to you, such that you are not currently engaging in widespread copyright infringement for using their artwork without permission or ownership…’
See, in the recent post asking for feedback on the weekend posts, one of the reasons I love the looking back on 5/10/15 years ago posts is how I get to be regularly reminded of where Thaler’s fool’s errand was at however many years ago, so I can laugh at him…
Does Thaler believe he got thrown under DABUS?
Re:
That’s cruel, but very usual PUNishment.
Another round of Monkeying around...
It’s one thing to shoot your own foot, something else entirely to call in a joint artillery and air strike on it.
When the law and legal precedent clearly says ‘Only human-created works get copyright’ and you base your entire case on how a human had no involvement in the creation of a work you’ve ensured defeat from the outset unless you were aiming to get even more courts to point to that sign.