US Copyright Office refuses to register AI-generated work, finding that "human authorship is a prerequisite to copyright protection"

Can a work entirely created by a machine be protected by copyright?

On Valentine’s Day, the US Copyright Office (Review Board) answered this question with a heartbreaking ‘no’, holding that “copyright law only protects “the fruits of intellectual labor” that “are founded in the creative powers of the [human] mind”” and consequently refusing to register the two-dimensional artwork 'A Recent Entrance to Paradise' below (the ‘Work’):

Creativity Machine's A Recent Entrance to Paradise


In 2018, Stephen Thaler (if the name rings an AI inventor-bell then you’re hearing it right) applied to register a copyright claim in the Work, indicating “Creativity Machine” as the author and Thaler as the owner of such machine.

The application stated that the Work had been autonomously created by a computer algorithm running on a machine. Registration was sought as a work-for-hire to the owner of the Creativity Machine.

In 2019, the Copyright Office rejected the application, holding that human authorship is necessary to support a copyright claim.

This conclusion is in line with what is expressly stated in the Compendium of Practices (p. 21-22):
[T]he 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. The crucial question is “whether the ‘work’ is basically one of human authorship, with the computer [or other device] merely being an assisting instrument, or whether the traditional elements of authorship in the work (literary, artistic, or musical expression or elements of selection, arrangement, etc.) were actually conceived and executed not by man but by a machine.” U.S. COPYRIGHT OFFICE, REPORT TO THE LIBRARIAN OF CONGRESS BY THE REGISTER OF COPYRIGHTS 5 (1966).
Thaler then requested a reconsideration of the decision, arguing that the human authorship requirement would be contrary to the US Constitution and be unsupported by either statute or case law. Such a request was once again unsuccessful.

A second request for reconsideration followed, also submitting that the Copyright Office “is currently relying upon non-binding judicial opinions from the Gilded Age to answer the question of whether [computer-generated works] can be protected.”

The Review Board’s decision

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.

AI-generated Kats
The Review Board 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.


The decision appears appropriate, also in light of international law.

While the Berne Convention does not define who can be regarded as an author, from its text and historical context, it appears that only natural persons who created the work can be regarded as authors. In particular, although Berne does not explicitly set an originality requirement, this already existed in national copyright laws at the time of drafting the Convention. According to Ricketson, it was clearly understood that this was also a requirement for the purposes of protection under the Convention, and inherent in the phrase ‘literary and artistic works’ in Article 2. The condition that a literary and artistic work possesses a sufficient (how much, however, the Convention does not say) degree of originality postulates "the need for the author to be a human being and for there to be some intellectual contribution above and beyond that of simple effort (‘sweat of the brow’) or what may be called mere ‘value in exchange’.’”

In any case, the above does not mean that there are no situations in which also works created by non-human authors can qualify for protection, a notable (and controversial) example being section 9(3) of the UK Copyright, Designs and Patents Act.

Looking at authorship from an EU perspective, Article 1(5) of the Sat-Cab Directive states that, for cinematographic or audiovisual works, the principal director shall be considered as its author or one of its authors, leaving Member States free to provide for others to be considered as co-authors. Article 2(1) of the Software Directive provides that the author of a computer program shall be the natural person or group of natural persons who has created the program or, where the legislation of the Member State permits, the legal person designated as the rightholder by that legislation. Article 4(1) of the Database Directive admits the possibility that the author of a database can be, not just the natural person or group of natural persons who created the base, but also—where the legislation of the Member States so permits—the legal person designated as the rightholder by that legislation.

In any case, the Term Directive refers the calculation of the term of protection of copyright to the life of authors as ‘physical persons’. In addition, the preamble to the DSM Directive specifies that the authors and performers that shall be able to rely on the provisions on contracts therein shall only be natural persons, thus excluding from the scope of application non-human authors and performers.

The Court of Justice of the European Union has not yet specifically tackled the question of who or what an author is. Nonetheless, it appears that its own understanding of originality – as a notion that presupposes a personal touch (Painer) and the making of free and creative choices (most recently, Brompton) – is indeed premised on the idea that authors in a copyright sense need to be human.

But the question that all romantics after Valentine's Day might ask is whether the last word on non-human authorship has been spoken yet. As it is the case of all real love stories, probably not. Thus: stay tuned for the next AI+IP development …

US Copyright Office refuses to register AI-generated work, finding that "human authorship is a prerequisite to copyright protection" US Copyright Office refuses to register AI-generated work, finding that "human authorship is a prerequisite to copyright protection" Reviewed by Eleonora Rosati on Thursday, February 17, 2022 Rating: 5


  1. Imagine being a student now and the questions we could ask in our exams. Fwiw I think this is ok for now, when AI takes over they'll clearly allow it ��. PS nice to see you all back, been following since the CII days in 2005

  2. The last word on AI authorship has not been spoken yet. Law will need to consider the extent to which AI/machine driven creativity should be recognised in authorship which in turn might involve a a consideration of the value of the initial human input in the product.

  3. You can get a copyright for a database, which often would not have an author in the conventional sense either.

  4. Granting copyright to an AI that was created and run by a person makes about as much sense as granting it to the paintbrush a painter used.

  5. Interesting! I don't interpret the findings in the Thaler cases the way that you have; in those cases Thaler's claim was specifically that the software had autonomously created the work, and since it was a "hireling" of Thaler's, Thaler should have the copyright. As you say, "Registration was sought as a work-for-hire to the owner of the Creativity Machine". (Thaler has similarly tried to get patents granted with his software as the inventor, and always failed similarly.)

    This is not at all the case with a typical work produced by a human using Midjourney or Stable Diffusion or Photoshop or whatever. In those cases, the human would be the author, and the work produced, not by the software or hardware acting autonomously, but by the human creator using the software and hardware; this is uncontroversial for Photoshop, and it would seem sensible for courts to rule the same in the case of AI tools. The human would therefore have copyright in the ordinary direct way, not through any work-for-hire complexity.

    One could make a case that the human isn't involved ENOUGH in the process of choosing an AI tool, deciding which of its engines to use if it has multiple, deciding on various settings, choosing a prompt, picking a particular produced image to preserve / display / etc. There are signs that the US copyright office is thinking about such things, but I don't know of any published finding or guideline.

    To me, the argument that a human is just as involved in producing a work with an AI tool as they are in taking a snapshot with a phone (something that uncontroversially results in copyright afaik) is a pretty strong one, along with the UK example of "whoever arranged for it to be produced"; but of course there's no predicting what the PO and the courts will decide!

    Thaler doesn't seem at all definitive on the issue, however. I've weblogged a bit about this; see and and update here:


All comments must be moderated by a member of the IPKat team before they appear on the blog. Comments will not be allowed if the contravene the IPKat policy that readers' comments should not be obscene or defamatory; they should not consist of ad hominem attacks on members of the blog team or other comment-posters and they should make a constructive contribution to the discussion of the post on which they purport to comment.

It is also the IPKat policy that comments should not be made completely anonymously, and users should use a consistent name or pseudonym (which should not itself be defamatory or obscene, or that of another real person), either in the "identity" field, or at the beginning of the comment. Current practice is to, however, allow a limited number of comments that contravene this policy, provided that the comment has a high degree of relevance and the comment chain does not become too difficult to follow.

Learn more here:

Powered by Blogger.