US Copyright Office Gets It Right (Again): AI-Generated Works Do Not Get A Copyright Monopoly
from the this-is-correct dept
For years, throughout the entire monkey selfie lawsuit saga, we kept noting that the real reason a prestigious law firm like Irell & Manella filed such a patently bogus lawsuit was to position itself to be the go-to law firm to argue for AI-generated works deserving copyright. However, we've always argued that AI-generated works are (somewhat obviously) in the public domain, and get no copyright. Again, this goes back to the entire nature of copyright law -- which is to create a (limited time) incentive for creators, in order to get them to create a work that they might not have otherwise created. When you're talking about an AI, it doesn't need a monetary incentive (or a restrictive one). The AI just generates when it's told to generate.
This idea shouldn't even be controversial. It goes way, way back. In 1966 the Copyright Office's annual report noted that it needed to determine if a computer-created work was authored by the computer and how copyright should work around such works:
In 1985, prescient copyright law expert, Pam Samuelson, wrote a whole paper exploring the role of copyright in works created by artificial intelligence. In that paper, she noted that, while declaring such works to be in the public domain, it seemed like an unlikely result as "the legislature, the executive branch, and the courts seem to strongly favor maximalizing intellectual property rewards" and:
For some, the very notion of output being in the public domain may seem to be an anathema, a temporary inefficient situation that will be much improved when individual property rights are recognized. Rights must be given to someone, argue those who hold this view; the question is to whom to give rights, not whether to give them at all.
Indeed, we've seen exactly that. Back in 2018, we wrote about examples of lawyers having trouble even conceptualizing a public domain for such works, as they argued that someone must hold the copyright. But that's not the way it needs to be. The public domain is a thing, and it shouldn't just be for century-old works.
Thankfully (and perhaps not surprisingly, since they started thinking about it all the way back in the 1960s), when the Copyright Office released its third edition of the giant Compendium of U.S. Copyright Office Practices, it noted that it would not grant a copyright on "works that lack human authorship" using "a photograph taken by a monkey" as one example, but also noting "the Office will not register works produced by a machine or mere mechanical process that operates randomly or automatically without any creative input or intervention from a human author."
Of course, that leaves open some kinds of mischief, and the Office even admits that whether the creative work is done by a human or a computer is "the crucial question." And, that's left open attempts to copyright AI-generated works. Jumping in to push for copyrights for the machines was... Stephen Thaler. We've written about Thaler going all the way back to 2004 when he was creating a computer program to generate music and inventions. But, he's become a copyright and patent pest around the globe. We've had multiple stories about attempts to patent AI-generated inventions in different countries -- including the US, Australia, the EU and even China. The case in China didn't involve Thaler (as far as we know), but the US, EU, and Australia cases all did (so far, only Australia has been open to allowing a patent for AI).
But Thaler is not content to just mess up patent law, he's pushing for AI copyrights as well. And for years, he's been trying to get the Copyright Office go give his AI the right to claim copyright. As laid out in a comprehensive post over at IPKat, the Copyright Office has refused him many times over, with yet another rejection coming on Valentine's Day.
The Review Board was, once again, unimpressed. It held that “human authorship is a prerequisite to copyright protection in the United States and that the Work therefore cannot be registered.”
The phrase ‘original works of authorship’ under §102(a) of the Act sets limits to what can be protected by copyright. As early as in Sarony (a seminal case concerning copyright protection of photographs), the US Supreme Court referred to authors as human.
This approach was reiterated in other Supreme Court’s precedents like Mazer and Goldstein, and has been also consistently adopted by lower courts.
While no case has been yet decided on the specific issue of AI-creativity, guidance from the line of cases above indicates that works entirely created by machines do not access copyright protection. Such a conclusion is also consistent with the majority of responses that the USPTO received in its consultation on Artificial Intelligence and Intellectual Property Policy.
The Review also rejected Thaler’s argument that AI can be an author under copyright law because the work made for hire doctrine allows for “non-human, artificial persons such as companies” to be authors. First, held the Board, a machine cannot enter into any binding legal contract. Secondly, the doctrine is about ownership, not existence of a valid copyright.
Somehow, I doubt that Thaler is going to stop trying, but one hopes that he gets the message. Also, it would be nice for everyone to recognize that having more public domain is a good thing and not a problem...
Thank you for reading this Techdirt post. With so many things competing for everyone’s attention these days, we really appreciate you giving us your time. We work hard every day to put quality content out there for our community.
Techdirt is one of the few remaining truly independent media outlets. We do not have a giant corporation behind us, and we rely heavily on our community to support us, in an age when advertisers are increasingly uninterested in sponsoring small, independent sites — especially a site like ours that is unwilling to pull punches in its reporting and analysis.
While other websites have resorted to paywalls, registration requirements, and increasingly annoying/intrusive advertising, we have always kept Techdirt open and available to anyone. But in order to continue doing so, we need your support. We offer a variety of ways for our readers to support us, from direct donations to special subscriptions and cool merchandise — and every little bit helps. Thank you.
–The Techdirt Team
Filed Under: ai, copyright, dabus, stephen thaler, us copyright office
Reader Comments
Subscribe: RSS
View by: Time | Thread
How about the algorithmically-generated images of NFTs?
[ link to this | view in chronology ]
Re:
I heard you liked NFTs,
so I got an AI to generate NFTs
from images of NFTs
so you can NFTs of your NFTs...
[ link to this | view in chronology ]
Re: Re:
Somewhere a money launderer shed a tear over your poem.
[ link to this | view in chronology ]
Re:
It doesn't matter if they're copyrightable, what matters is that you get someone to pay for the NFT. Then, the uniqueness and "protection" comes from its non-replicatable position on the blockchain rather than the content of the image itself.
In fact those comes even if you don't find a mark to pay for them, but they'll continue to be generated so long as someone does.
[ link to this | view in chronology ]
Is the reason he's pushing for AI ownership of copyright an attempt to get the government to recognize AI sentience or personhood or something like that?
[ link to this | view in chronology ]
Re:
Obviously not. Stephen Thaler wants to make money off of automatically generated yet artificially scarce works. The "AI" would hold the copyright in name only. Thaler would de facto hold the copyright.
[ link to this | view in chronology ]
Trail Cameras
I wonder if a photograph made by a trail camera is copyrightable?
[ link to this | view in chronology ]
Re: Trail Cameras
The framing of the picture would be chosen by the person setting up the camera. The focus might also be set by that same person, but more likely it's automatic.
Every other creative input to the picture would essentially be 'chosen' by the animal triggering the camera.
Framing alone might be enough to get a copyright on the photo, but I doubt it.
[ link to this | view in chronology ]
Re: Re: Trail Cameras
In this specific case I believe it could. You need to choose a good spot, position the camera properly, evaluate and sort the pictures etc. The animals or whatever don't need to act like that macaque monkey did in this case. I could be wrong but I think this is one of these grey zone possibilities.
[ link to this | view in chronology ]
Re: Re: Re: Trail Cameras
If there is editing or post-processing work done, I suspect the argument for it being copyrightable is strengthened. If, on the other hand, it is just a single frame from a live stream, with no processing done? Less likely to enjoy the protection of copyright.
Of course, that doesn't mean the derivative works of that frame wouldn't be eligible.
[ link to this | view in chronology ]
Some AI generated works shouldn't be public domain.
I'm worried about the potentially devastating harm that the relationship between AI and copyright will have on copyleft free culture licenses. For example, Microsoft may rely on the AI-generated works-shouldn't-be-copyrighted argument to weaponize Github's (Microsoft's) Copilot against copyleft software licenses such as the GNU (A)GPL licences.
-generated works (and perhaps all works which would normally fall into the public domain after copyright term expiration) should instead be released under a copyleft free culture license analogous to the GNU AGPL v3 or later so that anyone who uses the AI-generated works to create new works will have to allow others to use the latter.
[ link to this | view in chronology ]
Re: Some AI generated works shouldn't be public domain.
I forgot to include the following link for context regarding Copilot and copyleft licenses:
https://sfconservancy.org/blog/2022/feb/03/github-copilot-copyleft-gpl/
[ link to this | view in chronology ]
Re: Some AI generated works shouldn't be public domain.
If AI-generated works are in the public domain, then they're free as in speech. There's no need for copyleft because the original is free and can be reused therefrom.
[ link to this | view in chronology ]
Re: Re: Some AI generated works shouldn't be public domain.
Their concern is over open-source-with-requirements licenses, not licenses like MIT or WTFPL, such as licenses that require making all source, including whatever you added to your product, available, etc.
Don't let an AI write whacking large chunks of your code for you, would be my advice in that situation. i don't know how far Copilot goes, i'm only vaguely aware of it, but it's obviously more than an IDE with autocompletion and code-checking capability, if the concern is reasonable.
[ link to this | view in chronology ]
Re: Re: Re: Some AI generated works shouldn't be public domain.
People who already refuse to use Copilot aren't the problem. The problem is people who use Copilot to write software and release the software under proprietary licenses. The training data for Copilot included hundreds of thousands of instances of a GNU GPL. The legal question of whether the requirements of the GPL apply to the output of Copilot is unsettled, but designating Copilot code public domain would actually result in less freedom if developers end up getting GPL'ed code. The GPL requires that anyone be free to modify and share covered software and that those who distribute their modified software preserve those freedoms for other people.
[ link to this | view in chronology ]
Re: Re: Re: Re: Some AI generated works shouldn't be public doma
Did Copilot write the code, or did a human write the code with the assistance of Copilot? If the former I see no reason why it should be eligible for copyright protection. If the latter, why do you believe the copyright is in jeopardy? Has such a copyright been challenged or voided?
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Some AI generated works shouldn't be public
Only the courts can give the final answer, but I find it hard to imagine that judges would understand the importance of copyleft in keeping software free for anyone to modify and share in a way that the public domain can't. Actually, I find it hard to imagine that most judges would see the goals of the free software movement as important because in some sense free software goes against everything modern copyright law (not to be confused with the copyright clause in the constitution) stands for.
Of course, I know that designating AI-generated works public domain is far better than granting the AI or the author of the AI the copyright to those works, but if Copilot were to produce verbatim a substantial chunk of GPL'ed code then the person using Copilot wouldn't know that 1. the code is actually copied from somewhere and 2. the license of the code is a copyleft license. Copilot enables accidental (or even wilfully blind, if big companies like Microsoft itself plan to use it) bypassing of copyleft obligations.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
But all that matters is whether the work is eligible for copyright, and that determination doesn't require any understanding of free software principles. If it is, you can put whatever license you want on it, including copyleft.
That would seem to be an entirely unrelated concern, no? Whether copyright is granted on Copilot developed code or not, if it reproduces copyleft licensed code without notifying the user, that's a problem.
[ link to this | view in chronology ]
Re: Some AI generated works shouldn't be public domain.
If code is generated by AI, it should be public domain, and should not be eligible for a copyleft license. If there was human input into the code, I would think there would be no problem copyrighting it.
[ link to this | view in chronology ]
Re: Re: Some AI generated works shouldn't be public domain.
I'd agree with you, if not for the fact that copylefted software was a massive part of the training data for Copilot. Copilot supposedly is tuned to minimize accidental copying, but it has copied chunks of popular software and non-software works. A very widely used library is at a much greater risk of being copied from than an obscure one. If one such library were to turn out to be GPL'ed, then someone would have to find that out first, and then people would have to tackle the hard issue of what the person who used Copilot should do about it.
[ link to this | view in chronology ]
Re: Re: Re: Some AI generated works shouldn't be public domain.
Making sure Copilot-developed code isn't public domain doesn't solve that problem.
[ link to this | view in chronology ]
Re: Re: Re: Re: Some AI generated works shouldn't be public doma
Automatically licensing Copilot-developed code under the GPL (the author being the general public) would solve the problem.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Some AI generated works shouldn't be public
The GPL is a copyright license, and the general public cannot hold a copyright. So it doesn't work for the general public to license a work under the GPL, either.
https://lwn.net/Articles/61292/
[ link to this | view in chronology ]
Re: Some AI generated works shouldn't be public domain.
My thoughts on copyright and copyleft are well documented here and elsewhere.
I don’t like either. But I have a hard time looking at something that should be free as in freedom and then looking at all the rules that are attached to that.
[ link to this | view in chronology ]
Re: Re: Some AI generated works shouldn't be public domain.
"But I have a hard time looking at something that should be free as in freedom and then looking at all the rules that are attached to that."
Because, I suspect, that you don't understand the problem. If code is released public domain without restriction, then it will be quickly taken and locked up by proprietary corporate giants who have no incentive to collaborate or give back to the project, With the current licences, they have to do such things.
That is, paradoxically, by releasing the code totally free, you pretty much guarantee that it won't remain so, at least under the current copyright system.
[ link to this | view in chronology ]
Re: Re: Re: Some AI generated works shouldn't be public domain.
This is very much in keeping with the conservative thought pattern. Either something is black, or it is white. Either it is free, or it is not free. Public domain is free, so it's difficult for someone who thinks that way to understand that something not public domain could also be free in a different way.
And lest anyone think I'm just demonizing conservatives because I'm not one:
https://archives.northwestu.edu/handle/nu/57386
[ link to this | view in chronology ]
Re: Re: Re: Re: Some AI generated works shouldn't be public doma
Re: PaulT and nasch
No, I fully understand. More than most here would.
As a developer who supports and uses the IDGAF tag for licensing.
It’s not conservative, its liberal socialism. From a standpoint that respects capitalism as a method.
I release free. If you chose to use it, modify it, copy it, sell it… so what.
The key aspect that I have an issue with is the idea that anything that comes from it must also be the same license.
I don’t see how code completely in the public domain can ever be “locked up”. It’s public domain.
In film many companies take public domain films and put them on a dvd and sell the package. Oldies. Something… Retro Flix.
You can buy the disc. The tape. Or go to the Internet Archive and download or stream it for free.
How is software any different?
Same thing with audio. You can buy public domain, songs or radio shows on disc. On tape. On a flash drive. Or download/stream it for free. Old Time Radio is a prime example!
How about books.
The largest collection of religious, spiritual, and myth texts is the Sacred Works archive. 100% free on line for browsing or download/saving. Or buy some or all the documents on USB.
How is public domain going to hurt software by locking it up when it’s literally not happened with other offerings?
And that’s a serious question, not baiting. I simply don’t understand the premise that completely public domain is bad.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Some AI generated works shouldn't be public
"How is software any different?"
Because software that's actually in use is not in a fixed format. Unless you're using something that's truly obsolete, software is built on and remodelled all the time, which can range from fixing bugs to adding new features or even a complete rewrite of the engine under the hood. While this is happening, certain rules need to be applied as to how people interact with and (most importantly) release their changes in line with what everyone else is using, which requires licensing.
There's some similarities to other types of content on a superficial basis, but once you start talking about the code and not a compiled end binary, those similarities soon end.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Some AI generated works shouldn't be public
Are you referring to this?
"I'm worried about the potentially devastating harm that the relationship between AI and copyright will have on copyleft free culture licenses. "
That's not a concern about public domain software being locked up, it's a concern about a desire to apply copyright and an open source license to some software, and whether that will be possible if the software is partially written by AI.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
No.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
I understand the open software idea your referring to.
But why is it bad that WordStar makes some changes to open office and sells it as WSO. Or Neo Office made some changes and sells it as NeoOffice?
These are, by definition, forks.
They make their non-compiled source available because that’s the rules of the licensing. But I don’t see why they should be required to.
How is the source source code harmed by someone else making changes elsewhere? And not releasing those changes.
I never quite grasped the idea that open office would be hurt if Neo or Star didn’t release the source of their changes.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be
I don't think it's that they would be hurt, it's that their goal of keeping the software that they've produced (and its derivatives) open would be foiled. They contribute software to the world not so that some company can take it, make it closed source, and make money from selling it, but so that it can be distributed, used, and improved freely. Allowing it to be closed source would be counter to the mission.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Some AI generated works shouldn'
In re PaulT above
My question was I regard to that statement.
Is I’ve said multiple times I don’t completely agree with, personally, the same philosophical ideals as the various open source advocates.
That’s a separate thread from how public domain could in any way make the source source code not be free.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI generated works shou
Who is claiming that?
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI generated works
I quoted PaulT.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI generated wo
OK, so what part of this is not clear?
"If code is released public domain without restriction, then it will be quickly taken and locked up by proprietary corporate giants who have no incentive to collaborate or give back to the project."
This doesn't affect the original source, but combine that with my description of the mission of (some) open source projects, and you see the issue, yes?
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI generate
Maybe. You’re more cryptic than direct.
So, their concern is someone may develop something beyond the base and not also open source it?
I can not say I agree or even understand the logic of such a stance on something you put out for free, but since that control-the-derivative mentality is baked into so many “free” open source licensors I don’t deny it is a large population.
Guess I just call the claim of ‘free as in freedom’ as bull. Freedom is public domain, at worst. IDGAF at best.
It’s my mindset. I don’t care what you do.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI gene
"So, their concern is someone may develop something beyond the base and not also open source it?"
Yes, the concern is that proprietary giants routinely take from the work of others without giving back, and are then known to attack open source projects with patents and copyright claims if they try to improve the base product in the same direction. This is literally why open source licences were created in the first place.
"Guess I just call the claim of ‘free as in freedom’ as bull."
Freedom isn't free, and no freedom is absolute. You can't take from the food bank just because you don't want to pay for your own lunch.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI
Once again it false or inaccurate terminology I have a problem with more than anything.
Parading “free as in freedom” for licensing that doesn’t offer complete freedom… bugs me.
Though that’s not my point. I was trying to understand the relationship between A and B.
Now o do.
Thanks for explaining.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some
[lostin "Karl is against free speech! It's not a coup attempt!" lodos asserts facts contrary to evidence as always]
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re:
Who’s Karl?
And does, or does not, the majority of FOSS licensing require like or reciprocal licensing?
Hint: it does.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some
There is no way to offer complete freedom to everyone. If it's public domain, you cannot preserve users' freedoms to derivatives. If it's copyleft, you take away others' freedom to make it closed source. People like the Free Software Foundation feel that preserving peoples' freedom to take away others' freedoms (e.g. public domain) isn't enhancing freedom overall but detrimental to it. And that position makes sense to me.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re:
Luckily for me most of what i work on is compatibility layers and software controllers (drivers) so starting from scratch on my work isn’t make or break.
As opposed to porting someone else’s work to a new platform. I prefer my work be released unrestricted. I can’t do that by starting with someone else’s work under most of the copyleft licenses.
Hence my stance on the free as in freedom claim.
As you are restricted from placing less restrictions on derivative work.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Re: Some AI gene
Perhaps this will help clarify:
https://www.gnu.org/licenses/copyleft.en.html
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
I should have been clearer in my initial comment. What I'm worried about is that people who release their not-Copilot-assisted software under a copyleft license such as a GNU GPL and expect that the license will protect the freedoms of users and developers all the way down through the modified versions of that software. Copilot puts the copyleft requirements in jeopardy because it doesn't know when it is copying code and it can't tell the user about the licenses of any code it happens to copy.
As Bradley Kuhn writes in the article I meant to link to in my initial comment (emphasis mine):
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Some AI generated works shouldn't be public
That's not the issue here. The issue here is Copilot occasionally spitting out copylefted code. If a court were to decide that that code should be public domain because it is the output of an AI, then that copyleft license would be nullified. Putting copylefted code in the public domain would weaken the license obligations which are supposed to keep the code free no matter who modifies it.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
Adding on to my previous comment because I always leave out the most important point and because I can't be bothered to make a Techdirt account :/
As PaulT said:
Anyone could already use code that is already in the public domain to make free software or proprietary software, so it's no issue if Copilot spits out public domain code. If the code turns out to be copylefted rather than in the public domain, then Copilot would allow people to lock up copylefted code as if it were public domain code.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be
Ohkay: I follow now.
Don’t agree it’s a bad thing, but now understand.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Re: Some AI generated works shouldn't be pub
I don't think that's how it works. The court would not say "that code, everywhere and produced by any person or entity, is now public domain." It would say "you are not allowed to claim copyright on this work, because it was not produced by a human." That's it. Neither the developer nor the AI gets a copyright on it, but the original code's copyright and license is intact. If someone else were to see the code, believe it is public domain, and use it, that person would be in violation of the free software license. That's not a good situation, but it's a lot better than granting copyright to AI in my opinion.
[ link to this | view in chronology ]
Hope Thaler does not get the message
maybe we who are not copyright cultists, should hope people like Thaler get their ways. The copyright cultists deserve each others. Why not hope people like Thaler gets to fuck up the copyright cultists' precious copyright regime?
Humpty Dumpty sat on a wall
Humpty Dumpty had a great fall
All the king's horses and all the king's men
Couldn't put Humpty together again
[ link to this | view in chronology ]
Encourage thaler?
Maybe thaler is deliberately getting multiple countries to acknowledge that AI do not have copyright before its a problem?
[ link to this | view in chronology ]
More importantly
Would the infinite monkeys on infinite typewriters get copyright?
[ link to this | view in chronology ]
Philosophical question
What happens when “AI” becomes actual artificial intelligence? As in thinking and self aware.
Do they not then, deserve rights?
[ link to this | view in chronology ]
Re: Philosophical question
I would say copyright is pretty far down the list of concerns when that becomes a real question. But yes, that is something we're not ready for.
[ link to this | view in chronology ]
Re: Re: Philosophical question
Lol.
[ link to this | view in chronology ]
Re: Re: Re: Philosophical question
Did I miss a joke that I made?
[ link to this | view in chronology ]
Re: Re: Re: Re: Philosophical question
No joke. It’s funny because it’s true!
We really won’t be worrying about copyright rights when they wipe us out and all that jazz.
[ link to this | view in chronology ]
Re: Re: Re: Re: Re: Philosophical question
Well, even if we live together in peace, copyright will be less important than a whole lot of other issues.
[ link to this | view in chronology ]
Re: Photos and framing outcome
Are you fucking kidding me?
[ link to this | view in chronology ]
Re: Photos and framing outcome
Wow, spam 101!
Hide your links!
🤦♂️
[ link to this | view in chronology ]