The Volokh Conspiracy
Mostly law professors | Sometimes contrarian | Often libertarian | Always independent
No Copyright in AI-Created Works, Says Copyright Office

From a Copyright Review Board decision (Re: Second Request for Reconsideration for Refusal to Register A Recent Entrance to Paradise) last week:
On November 3, 2018, Thaler filed an application to register a copyright claim in the Work. The author of the Work was identified as the "Creativity Machine," with Thaler listed as the claimant alongside a transfer statement: "ownership of the machine." In his application, Thaler left a note for the Office stating that the Work "was autonomously created by a computer algorithm running on a machine" and he was "seeking to register this computer-generated work as a work-for-hire to the owner of the Creativity Machine." …
[C]opyright law only protects "the fruits of intellectual labor" that "are founded in the creative powers of the [human] mind." U.S. Copyright Office, Compendium (Third) of U.S. Copyright Office Practices § 306 (quoting Trade-Mark Cases (1879)); see also Compendium (Third) § 313.2 (the Office will not register works "produced by a machine or mere mechanical process" that operates "without any creative input or intervention from a human author" because, under the statute, "a work must be created by a human being"). So Thaler must either provide evidence that the Work is the product of human authorship or convince the Office to depart from a century of copyright jurisprudence. He has done neither.
Thaler does not assert that the Work was created with contribution from a human author, so the only issue before the Board is whether, as he argues, the Office's human authorship requirement is unconstitutional and unsupported by case law. {Because Thaler has not raised this as a basis for registration, the Board does not need to determine under what circumstances human involvement in the creation of machine-generated works would meet the statutory criteria for copyright protection. See Compendium (Third) § 313.2 (the "crucial question" of human authorship is whether a computer is "merely being an assisting instrument" or "actually conceive[s] and execute[s]" the "traditional elements of authorship in the work").} Currently, "the Office will refuse to register a claim if it determines that a human being did not create the work." § 306. Under that standard, the Work is ineligible for registration. After reviewing the statutory text, judicial precedent, and longstanding Copyright Office practice, the Board again concludes that human authorship is a prerequisite to copyright protection in the United States and that the Work therefore cannot be registered.
The Copyright Act affords protection to "original works of authorship" that are fixed in a tangible medium of expression. 17 U.S.C. § 102(a). The phrase "original work of authorship" … is "very broad," but its scope is not unlimited. Congress chose this language to encompass a smaller set of creative works than could be protected under the Constitution. Because of this gap, the Act leaves "unquestionably other areas of existing subject matter that this bill does not propose to protect but that future Congresses may want to." …
[C]ourts have repeatedly rejected attempts to extend copyright protection to non-human creations. For example, the Ninth Circuit held that a book containing words "'authored' by non-human spiritual beings" can only gain copyright protection if there is "human selection and arrangement of the revelations." Urantia Found. v. Kristen Maaherra (9th Cir. 1997) (holding that "some element of human creativity must have occurred in order for the Book to be copyrightable" because "it is not creations of divine beings that the copyright laws were intended to protect"). Similarly, a monkey cannot register a copyright in photos it captures with a camera because the Copyright Act refers to an author's "children," "widow," "grandchildren," and "widower," — terms that "all imply humanity and necessarily exclude animals." Naruto v. Slater (9th Cir. 2018); see also Kelley v. Chicago Park Dist. (7th Cir. 2011) (rejecting a copyright claim in a "living garden" because "[a]uthorship is an entirely human endeavor" and "a garden owes most of its form and appearance to natural forces"). These court decisions are reflected in the Office's guidance in the Compendium, which provides examples of works lacking human authorship such as "a photograph taken by a monkey" and "an application for a song naming the Holy Spirit as the author." While the Board is not aware of a United States court that has considered whether artificial intelligence can be the author for copyright purposes, the courts have been consistent in finding that non-human expression is ineligible for copyright protection.
{It appears, however, that the Eastern District of Virginia held that artificial intelligence systems cannot claim inventorship of patents. Thaler v. Hirshfeld (E.D. Va. 2021) (concluding that under the Patent Act, "an 'inventor' must be a natural person" and upholding refusal of a patent application). Though the court's opinion was based on construction of the Patent Act, rather than the Copyright Act, the similarity of the court's statutory analysis to that in the copyright cases relied on by the Board supports the conclusion here.} …
Thaler's secondary argument, that artificial intelligence 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, is similarly unavailing. First, the Work is clearly not a work made for hire as defined in the Copyright Act. A work made for hire must be either (A) prepared by "an employee" or (B) by one or more "parties" who "expressly agree in a written instrument" that the work is for-hire. In both cases, the work is created as the result of a binding legal contract—an employment agreement or a work-for-hire agreement. The "Creativity Machine" cannot enter into binding legal contracts and thus cannot meet this requirement. Second, the work-for-hire doctrine only speaks to the identity of a work's owner, not whether a work is protected by copyright. As explained above, the statute requires that a work contain human authorship. In sum, the Work is not a work made for hire because it is neither a "work of authorship" nor a work created "for hire."
Much of Thaler's … request amounts to a policy argument in favor of legal protection for works produced solely by artificial intelligence. He cites to no case law or other precedent that would undermine the Office's construction of the Copyright Act. Because copyright law as codified in the 1976 Act requires human authorship, the Work cannot be registered….
Thanks to the Media Law Resource Center MediaLawDaily for the pointer.
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time. Comments may only be edited within 5 minutes of posting. Report abuses.
Please
to post comments
Under Copyright Law, copyright vests originally in an "author," although it can be assigned later.
Who in an AI case is the author? Can a computer progam own a copyright, or any other property?
As the decision above explains, there's no copyright to vest without "the creative powers of the [human] mind". Therefore, you never get to the question of authorship.
Shouldn't the person who designed the AI be the 'author' for all derivative works? They designed the AI to do what it does, it's just a tool they're using to do it.
They could have claimed that, but they chose not to, apparently trying to make some point about non-human authorship.
That was the chain of reasoning I went through, that they probably could have succeeded on that claim, but were likely trying to mildly lawfare up some precedent for future AI developments or something.
So companies that make paper used in books should get copyright to any books made with their paper?
More on point, I saw a fellow boasting on Twitter that he created an algorithm to solve Wordle in as few turns as possible. If the algorithm solves wordle in one try, would you credit its creator with solving wordle in one try?
If the algorithm works, then the algorithm works.
there's people out there who have an algorithm to solve Rubik's cubes. Whoever first developed the algorithm gets credit for solving Rubik's cube.
As several people have pointed out, that would be a plausible claim but it was specifically not made in this case.
That said, I think that hypothetical claim should fail. Copyright derivative use is generally analyzed under the Fair Use doctrine. And that's as a derivative of other copyrighted material. But the AI is not copyrighted - it is invented and thus patentable. The analysis there is different.
Consider a fancy handbag-making machine. You own the patent and have used it to make only one machine. You can use that to keep other people from copying your machine but your patent won't protect you from people copying the handbags (using other machines or technologies). You could separately protect the design of your fancy handbags but you don't automatically get that protection as a derivative from your patent.
For a maybe-closer analogy, you do not violate Mirriam-Webster's copyright when you use their thesaurus to write a poem. There is no automatic 'derivative authorship' or copyright.
If I create an automated system that can create otherwise copyrightable works, then those works are derivative of the original work (creating the automated system.) 17 USC 106 expressly grants exclusive rights in derivative works.
"Under Copyright Law, copyright vests originally in an "author," although it can be assigned later.
Who in an AI case is the author? Can a computer progam own a copyright, or any other property?"
to answer your later questions, I suggest referring to law in the 1850's. If a black man wrote a book, and that black man happened to be an American slave, would the owner of the slave be able to claim copyright? A variation of "work for hire" perhaps?
This seems wrong to me. If an author uses a word processor with spelling and grammar checking, does that make it uncopyrightable?
If the author uses a paper thesaurus and dictionary, does that make the result uncopyrightable?
An AI is just a different tool.
On the other hand, IANAL and don't understand the legal claim:
Is he claiming to be the author or not? Why drag the tool into the matter? That adds some legal quibblery which leaves me unsure of exactly who is supposed to get the copyright.
He was claiming that the tool is the author, which is why the Copyright Office denied his registration and his requests for reconsideration. He probably wanted to make history by having his software be the first non-human to be registered as the author of copyright-covered material, which is why he cares about the tool being listed as the author.
Yes.
It seems likely that one or more humans could have registered a copyright in this work. These humans may have, for example, exercised creativity in selecting training sets to train the AI program, provided creative guidance to the program by setting various parameters, and exercised creativity in discarding the other 1e12 generated images that were complete and utter garbage (or would violate pornography laws in some places).
Moral of the story: if you want a copyright, don't be such a tool.
Hire some lobbyists instead. Disney got the Mickey Mouse Copyright Term Extension Act passed.
"Thaler does not assert that the Work was created with contribution from a human author"
They addressed your point in the opinion. It's not that the work was created with partial use of a machine, it's that the machine on its own, without any human contribution, created the work.
I mean, that was asserted, but obviously it's not true; someone had to design the machine and tell it what to do. Those are human contributions.
It wasn't an attempt to obtain a copyright, it was an attempt to set a legal precedent. If they'd wanted the copyright itself, they would have claimed it directly, asserting the AI was a tool used in the creation of the work.
Which it obviously was.
" it's that the machine on its own, without any human contribution, created the work."
This would be so, if anyone had ever built a machine capable of spontaneously creating literary works. But nobody has ever done this. Shakespeare wrote "Romeo and Juliet". Countless hacks have written variations. George Lucas used to make "Star Wars" movies. Then Disney bought the brand, and hired JJ Abrams to make movies that were branded as "Star Wars" movies, but which were not. Then they hired Dave Filoni to make some "Star Wars" TV shows. Those projects succeeded. In both cases, the new creators had access to all the original films (in all their variations.)
An AI can do the same thing... if you program it with how to build a story, and load in some starting point, it can build you a story. Depending on how well the AI was built, the story you get out the other end may be a "Force Awakens" or may be a "Mandalorian".
That is a good point which I did not think about. If a machine can own a copyright, can it get civil rights and standing as well?
The intent here seems to be specifically to open that can of worms. Of course, by denying the copyright we may have opened a completely different can of worms
Whether or not a machine can gain civil rights is not that different from whether or not a person with darkly-colored skin can. Before the Civil War, the notion that black people were people was as ludicrous to jurists of the time (Taney) as the idea that machines might be people is to jurists of present day.
In order to qualify for copyright, you need some level of creativity, even minimal. In the examples you give, a human being supplies the creativity, and may use those mechanical tools to help him or her. The copyright covers on the creative part, not the proper spelling or grammar.
Disney gained copyright protection for "The Force Awakens", despite the fact that the only creativity applied to producing it consisted of "let's pretend this isn't a remake of the 1977 Star Wars movie".
For that matter, the remake of Psycho got copyright protection despite the fact that it is the same movie as the original Psycho, except with different faces. Looping back to Star Wars, but this time to Rogue One, in that film some of the faces were NOT changed, despite the fact that the owners of those faces had made changes.
If you type a word to seed the computer and keep accepting autocompletion results, I don't think you have put any meaningful creative work into it. If you let the computer normalize spelling the creative work is yours.
"If you type a word to seed the computer and keep accepting autocompletion results, I don't think you have put any meaningful creative work into it"
But has whoever wrote the autocomplete program? Or the crowd the autocomplete draws from?
As others have pointed out, in your examples the author is providing the creative drive...whether word processors or thesauruses are used to assist him or her doesn't make a difference.
In the above case, the "owner" is asserting he provided zero creative impetus to the manuscript. Therefore they don't get copyright, anymore than Microsoft would get copyright over a manuscript you used MS word to make.
"As others have pointed out, in your examples the author is providing the creative drive...whether word processors or thesauruses are used to assist him or her doesn't make a difference."
Counterpoint: Jabberwocky. If you run that through a spelling-and-grammar checker, you are going to get a completely different result from what Mr. Dodgson created.
Counterpoint #2: The Xanadu poem that was interrupted by the neighbor from Porlock. Was that creativity from a human mind or from the author's drug use? Does pursuing this inquiry suggest a different outcome in deciding the authorship of "Lucy in the Sky with Diamonds"?
Does automated recoloring of a human-created input picture à la NFTs have enough human creativity to be copyrighted?
How about when a news program takes surveillance video and uses it to show a news event? Is the copyrightable element the setting up of the camera and recording equipment, or the selection of which frames to broadcast?
I think if the author set an AI program in motion with some selection of parameters then that is a creative decision and enough to establish authorship.
If however all that was done is to start a program with no human interaction then there is no creativity involved and no copyright.
However if a person created an AI and used it to create images then there may be some creativity involved in the creation of the tool and the resulting product.
You can copyright the program. Copyrighting what the program produces....that's more tenuous.
Can be done. In the case of code generating computer screens, both the code itself, as well as the screens generated, have had their copyrights withstand legal attack. Or, in one case I worked on, the code in two programs had no real similarity, but the second was found to have been copied from the first by the striking similarity of the screens.
Machines are 100 times better than living beings at any task. This decision seems unfair and self defeating. It will not protect a superior product. The copyright law should be changed to allow a product from any source.
The idea of the law is encourage innovation. So, lawyer, dumbass, which term will encourage innovation more, a 5 year term or a 100 year term? Think, for once, think.
Whoever wrote your program deserves life imprisonment. Your code is flaky.
Workaround?
The AI program is copyrightable. Add a random number generator, so that with each [unique] random seed, a unique work is produced.
The human determines the seed, so the work is now the product of a human mind. Copyrightable? Even if they reproduce the AI, they cant reproduce the work without the human-determined seed.
Anyone remember the monkey selfie case?
I was thinking of the same thing. Here and in the monkey case the problem for the owner of the hardware is it is the alleged non-human authorship that makes the work saleable. You could call it "art" or "a picture of a monkey" and remain silent on the details, and nobody would pay attention.
Sure. The monkey did not take the picture.
With equipment which exists now, the monkey could take that picture. With the equipment which existed then, it was impossible.
"Sure. The monkey did not take the picture. "
If you can modify the hypothetical by altering the facts, you can get any result you like. In the monkey-selfie case, the monkey DID take the picture.
The monkey took the picture. the problem is "can you show that the monkey MEANT to take the picture?", to which the answer was no.
Terrible decision. Did not address or even cite to the precedent set in “Author, Author,” (ST: VOY s7e19).
https://memory-alpha.fandom.com/wiki/Author,_Author_(episode)
Star Trek as reference is sloppy at best. The timeline(s) keep shifting.
If you want solid precedent, you want to go to "Jerry Was A Man", Robert A. Heinlein, circa 1941.
Thank you, this makes me wonder why we even bother having courts.
Machines and programs will only ever do what their programmers and operators enabled or directed them to do, intentionally or not. They're not individuals or conscious. But there's of course a monumental cultural obsession with a fantasy to the contrary.
Viruses and bacteria only do what their genetic programming tells them to do. At what point are animals more than the sum of their genetic code?
Once machines are self replicating and can manipulate their own code, we are beyond what the programmers and operators enabled or directed them to do.
A self replicating machine that adapts to its environment has a rudimentary level of "consciousness."
Now if I take a self replicating machine and give it the ability to change its own programming, we have achieved a new layer of consciousness. In the same way I can program a machine learning algorithm with art to produce new art, I can program a machine learning algorithm to accepts the code of all the other machine learning algorithms out there, then spits out another one modulo some randomness, similar to they way this AI art works.
Life on planets without atmospheres, could easily be complex "machines."
And so, the VC commentariat approacheth Aristotle.
"Machines and programs will only ever do what their programmers and operators enabled or directed them to do, intentionally or not. They're not individuals or conscious."
Biology is machinery. People have been creating conscious individuals via biology for thousands of years. They're called "children."
Do you remember the old Bob Newhart routine, where someone decides to test the proposition that 10,000 monkeys pounding away at random on 10,000 typewriters would eventually create the complete works of Shakespeare?
"Hey, Chief, I think we may have something here. Let's see. 'To be, or not to be. That is the gazornenplatz.'"
"gazornenplatz" -- Is that Danish for existential question?
I forget who suggested that if 10,000 monkeys, allowed to use typewriters for as long as necessary would eventually produce Shakespeare, but the smell would be horrible.
Arguably, human beings likewise only do what their genetic programming and environmental stimuli in combination tell them to do, and so are just like machines. The consequences of such a position would seem to include that humans do not have any free will and are not morally accountable for anything (assuming moral truth even exists). Most people intuitively reject this, although excusing oneself of moral accountability holds a certain attraction.
On the other hand, there is a fantasy of a machine or program having its own consciousness and free will. This is the other side of the same coin, I think -- rejecting the existence of free will, consciousness and moral accountability altogether, is basically similar to grossly distorting those concepts in such a way that they apply equally to machines and viruses and bacteria. This is the coin that people seem to overwhelmingly reject on an intuitive level as not corresponding to reality, even while many maintain a sort of morbid fascination with the idea.
Not sure why you would say that "replication" combined with response to environmental stimuli necessitates "consciousness" under any sort of meaningful definition. Any computer program or machine can do this. Computer viruses and biological viruses and bacteria all do it, yet are not conscious. Siri on your iPhone on the other hand, is ostensibly aware of its own existence, name, purpose, and such, and so could be deemed "conscious" under some definition, but again it does not seem like a very good definition. Ask Siri about its self awareness and it will have all kinds of thoughtful and good responses, but those aren't Siri's thoughts, they are just scripts written by the programmers.
"give it the ability to change its own programming" Seems like a non-sequitur. Any such changes are simply a carrying out of the original programming. Any simple computer program can be designed to "manipulate" its own code.
^meant to reply to dwb above
Could the AI sue for violation of its copyright? And if it won the suit, would the defendant have to pay it in bitcoin?
"Your honor, I call Creative Machine to the stand."
Futurama did an episode on whether or not artificial intelligences have or can exercise free will. And we all have the collected warnings of Skynet in the Terminator movies.
I know I am a day late to the party... but there is a chance someone will come across this and find it funny. This all reminds me of a simple poem someone once told me as they were helping me fix my computer.
I really hate this stupid machine.
I really wish we'd sell it.
It never does what I want,
But only what I tell it.
Did you just violate somebody's copyright?