US Court of Appeal confirms human authorship requirement, including for AI

A Recent Entrance to Paradise

Can a work that is presented as being entirely AI-generated receive protection under copyright law? The answer, unsurprisingly, is ‘no’.

Earlier this week, the US Court of Appeal for the District of Columbia Circuit upheld all previous decisions that had refused to accept copyright protection under the 1976 Act for A Recent Entrance to Paradise, an artistic work that is claimed to have been entirely generated by Dr Stephen Thaler’s Creativity Machine.

By way of recap, in 2019 Thaler filed an application with the Copyright Office to register the work, and indicated the Creativity Machine as the sole author and himself as the work’s owner. The Copyright Office rejected the application. The refusal was subsequently upheld by the Review Board first [IPKat here] and then by the US District Court for the District of Columbia. Before the latter, it may be recalled that not only did Thaler argue that he would be the owner of the copyright in the work by virtue of the work-for-hire doctrine, but also that the work would be protected because he “provided instructions and directed his AI”.

Now, the Court of Appeal has confirmed that “the Copyright Act of 1976 requires all eligible work to be authored in the first instance by a human being.” Incidentally, the Copyright Office expressly spelled out such a requirement as early as 1973, with reference to the then emerging employment of computers.

The decision

While the statute does not define ‘author’, this notion only refers to human beings “[because many of the Copyright Act’s provisions make sense only if an author is a human being”. This would be confirmed by the following:

• The ownership provision is premised on the author’s legal capacity to hold property.
• The duration of protection is linked to the author’s life.
• Upon the author’s death, rights can be exercise by surviving spouses, children and grandchildren.
• Copyright transfers require a signature.
• Authors of unpublished works are protected irrespective of their nationality or domicile.
• Authors have intentions, including that of working together to create works of joint authorship.
• The Copyright Act refers to machines as tools, not as authors.

While not all legally recognizable authors fulfil the requirements above (think of people without a family or domicile, or the fact that married women in the 19th century could not own property, let alone copyrights), machines would plainly have no legal capacity, life expectancy, families, the ability to sign documents, nationality, domicile, or intentions.

No authorship of stray Kats: it is because they 
are not human or because they have no domicile?

As a result, “[t]he human-authorship requirement, in short, eliminates the need to pound a square peg into a textual round hole by attributing unprecedented and mismatched meanings to common words in the Copyright Act.”

That said, the Court noted – and this is key – that “adhering to the human-authorship requirement does not impede the protection of works made with artificial intelligence.” That is so inter alia because a human can create works with the assistance of AI and those works can receive protection. The thing with Thaler was that he indicated the Creativity Machine as the sole author of the work.

Finally, the Court was unpersuaded that a requirement of human authorship would disincentivize creativity by creators and operators of AI.

Comment

As noted in the opening, the decision is unsurprising and broadly correct. The real ‘problem’ is that Dr Thaler’s case, a few years on, is no longer particularly interesting or relevant.

The reason for such a ‘harsh’ judgement is twofold: first, that he listed the Creativity Machine as the sole of author of A Recent Entrance to Paradise and, second, that some of the arguments raised (including alleged contrariety to the US Constitution of a human authorship requirement and that Thaler himself is the author of the work) have not been pursued further.

The real question is not “who” an author can be, but rather “what” makes one an author.

Put otherwise: how far can one push the use of AI or other machines and still be regarded as the author of the resulting output? When we speak of ‘authorship’ or also, by reference to the originality requirement, ‘personal touch’ and ‘freedom and creativity’ of the author, what do we mean?

These are necessarily questions of fact and degree, but – in my view – they are the only meaningful ones in the context of AI ‘authorship’ [see also here and here].

Indeed, if look at US practice, the recently released Part 2 of the US Copyright Office Report on ‘Copyright and Artificial Intelligence’ confirms that the use of AI tools to assist human creativity is no bar for copyright protection of the output insofar as there is sufficient human control over the expressive elements.

Unlike what some early commentators argued, this is not at all a change of position on the side of the Office. It is instead perfectly consistent with the Compendium of Practices itself (and the 1965 Report to the Librarian of Congress), which states (at §3131.2, emphasis added) that “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”.

And this is the reason why the recently granted registration for A Single Piece of American Cheese [IPKat here] is neither surprising nor incorrect. The applicant argued the case well and provided appropriate evidence of Kent Keirsey’s creative input.

And what about Europe? Well, things are not that dissimilar from the US. 

The arguments presented by the US Court of Appeal to note the requirement of human authorship under the 1976 Act are broadly transferrable to the European context as well, thus suggesting that also here there is a requirement of human authorship. 

In any event, the originality requirement, shaped through years of case law of the Court of Justice of the European Union (CJEU), further mandates that the ‘personality’ of the author is visible in the work, which must be the result of their “free and creative choices”. It thus follows that, under EU law too, works can be generated with the assistance of AI and other machines, but what is going to be protected is solely the parts in respect of which the (human) author was able to exercise the required originality. Still through the CJEU we know that choices dictated by technical considerations, rules and constraints do not confer protection, at least under copyright law. 

When the CJEU decides Mio [IPKat here and here] and konektra [IPKat here] – the cases have been joined and the Opinion of Advocate General Szpunar is expected in early May – as well as Institutul G. Călinescu [IPKat here], important additional pieces of the puzzle might be added.

In conclusion, the last word on AI-assisted works is still a long time away from being spelled out. Hopefully, future decisions will help shed light, not on the (now) irrelevant question of whether only humans can generally be regarded as authors, but rather on the key issue of what authorship and originality entail. And this is a question that goes to the core of copyright law, well before, well beyond and well after AI.