AI in the Creative Sector: Brushing Up on Copyright for AI-Generated Content

There's a lot of noise in the market right now when it comes to artificial intelligence. It's true, that the legal issues arising in this context are multiple and complex—especially when we venture into the realm of intellectual property. However that doesn't mean that the issues are insurmountable; just that we should be prudent as we embrace the latest in a long series of new technologies.

There's no doubting that the creative industries are facing a variety of challenges when it comes to developing, deploying, and using AI. Not least that how much and how quickly the status quo will be disrupted remains unclear.

Legal challenges are playing out in courtrooms around the world, ranging from debate as to whether 'scraping' data to train AI is an infringement of intellectual property rights, to whether AI-generated content might infringe upon the personality rights of performers. Whilst we are advising clients on those questions, in this article we'll focus on the issues being debated about subsistence of copyright in AI-generated content, particularly in Europe, the UK, and the US.

Who owns the output of AI?

Assuming there is anything capable of copyright protection at all—which we'll come to in a moment—who owns the copyright is an important question having economic impact and dictating subsequent transaction structures.
Where multiple people work together to create content, or where one person engages another to create content, it will be a matter of contract as to who owns the copyright in the output. In the absence of a valid contractual assignment or employment relationship, the ownership will remain with the person that the law provides is the first owner.

In the EU, the author who creates the work will be the first owner of copyright.

In the UK, this issue is addressed by the Copyright, Designs and Patents Act 1988 ("CDPA"). The CDPA anticipates "computer generated works" as being capable of copyright protection, with the first owner being the person who made the "arrangements necessary for the creation of the work".

In the US, the relevant legislation is the Copyright Act, with the US Copyright Office administering the copyright registration system. Whilst registration is not necessary for copyright to subsist, it is required for the right to be enforced in the US. Again, the first owner of a copyright work will be the author, subject to any contractual arrangements.

As far as contractual arrangements go, in most cases, the terms of use of AI platforms provide that the user is the owner of (and responsible for) the output.

Is the output capable of copyright protection?

The answer to this question depends on the relevant jurisdiction.

EU Legal Position

In the EU, there will need to be a human author for copyright to subsist in the works generated.
Certainly in the Infopaq[1] case, whilst not considering AI specifically, the European court decided that for a work to qualify for copyright protection, it must be "the author's own intellectual creation". It clarified that a degree of creativity (i.e., from a human) is required to establish copyright protection, and this decision has since been followed by later judgments.

Nevertheless, this interpretation would not prevent copyright from subsisting in a work where an AI has been used to assist in the creation, provided of course that there is sufficient human input to have satisfied the intellectual creation test. The European court helpfully highlighted in Eva-Maria Painer v Standard Verlags GmbH[2] that it is possible to create works capable of copyright protection using computer-based tools. The debate here will likely centre around determining to what extent human involvement is required.

It's critically important to recognise that there are likely to be rapid developments in this area, as we were reminded when writing this article by President Macron's party introducing its proposal to the French National Assembly aimed at copyright in the context of AI, which proposes that the first owners of a work created by AI without human intervention are those whose works were used to create the new content.

UK Legal Position

Under English law there is no express requirement for the work to be the intellectual creation of a human author to qualify for copyright protection. Instead, authorship is granted to the person making the arrangements necessary for the creation of the work. The lack of express requirement for a human author and narrow statutory wording under the CDPA allows for arguments to be made in favour of granting copyright protection even for purely AI-generated works.

Rather than whether copyright subsists in a machine-generated work, the debate in the UK is more likely to revolve around determining who can be said to have made the arrangements for its creation. If questions about intellectual creation, skill, or labour are going to arise, this could be the area in which we see them argued.#

US Legal Position

The US Copyright Office has made it clear that purely AI-generated content, not having a human author using a degree of creativity, does not enjoy copyright protection (and therefore there is no 'author' for copyright purposes). This has been evidenced by various cases, such as Steven Thaler's AI-generated painting 'A Recent Entrance to Paradise', which was denied copyright protection in the US due to the lack of human authorship.
There are suggestions that if a human contributes significant creative effort to the final AI-generated content, that content may be considered copyrightable by the US Copyright Office—in an approach reminiscent of the EU position. The key here is that a human author must have crafted something beyond the AI's direct output to produce a unique piece of work. In that case, the final product produced by the artist might be eligible for copyright protection under US law. It strikes us that, as is often the case, we're again going to be dealing with a grey area where the threshold degree of human input remains to be determined, and it's reasonably likely that a different threshold will apply in the US compared with the EU.

There is some comfort for AI-enabled creators however. Whilst a notable case in 2022 involving Kris Kashtanova's comic book 'Zarya of the Dawn' further emphasised the rule, with the US copyright office deciding that the AI generated images in the book were not eligible for copyright, the case did highlight the potential for copyright to subsist in the arrangement of works (which individually may or may not be copyright protected). Indeed, Kashtanova obtained copyright protection for elements such as the (human-authored) writing, and the order in which the AI-generated images appeared in, but not the images themselves.

In any case, it must be noted that copyright applicants in the US must disclose the use of AI in their registrations, and existing applications without this information need to be amended.

The Berne Convention

One way in which copyright protection may be enforced internationally is through the Berne Convention, to which the EU members, the UK, and the US are signatories. This is based on essentially three basic principles:

• works originating in one of the contracting states—that is, works the author of which is a national of such a State or works first published in such a State—must be given the same protection in each of the other contracting states as the latter grants to the works of its own nationals;

• protection must not be conditional upon compliance with any formality; and

• protection is independent of the existence of protection in the country of origin of the work. The extent of protection and remedies available in respect of infringement will depend on the local laws of the state in which the infringing act occurs.

In simple terms, if the US courts refuse to protect purely AI-generated works but the UK does not, protection may nonetheless be available in the UK for works originating in the US. That work will not, however, be protectable in the US.

Practical Implications

The variation in treatment for copyright around the world (taking the EU, UK, and US as examples) is going to present a challenge for the use of AI in the creative industries. These are powerful tools that we should be embracing where it's going to add value, but protecting that value is a challenge that needs to be addressed before risking substantial investment.

Whilst works can be protected under arrangements between contracting parties without the need for IP to subsist, this becomes problematic when that work is to be communicated to the public. This represents a substantial challenge for the creative sector where the purpose of creation is to deliver content to the public, if there is no privity of contract with the end user who may be able to reproduce, modify, and distribute works freely without risk of legal liability. We will therefore need to be creative when it comes to protecting our investment.

The rules relating to AI and copyright are intricate and subject to a developing area of law. Fieldfisher have been advising clients on this topic and have extensive experience in this area of law. If you would like any advice on this topic, please contact Chris Eastham (chris.eastham@fieldfisher.com), Tim Johnson (tim.johnson@fieldfisher.com), Stephen Saltzman (stephen.saltzman@fieldfisher.com) or your usual Fieldfisher contact.

Thanks to Arwa Abdeh, solicitor in Commercial IP, for her assistance in creating this article.