Do Androids Dream of Electric Copyright? This allusion to the title of Philip K Dick’s 1968 dystopian novel, on which the 1982 film Blade Runner was based, is the playful title of a scholarly paper about authorship of computer-generated art. Published in 2017 and written by Andrés Guadamuz, reader in intellectual property law at the University of Sussex, the discourse prefigures practical concerns now emerging in the contemporary art world surrounding artificial intelligence (AI). Disquiet is currently developing following the March 2023 publication by the US Copyright Office of instructive guidance: Works Containing Material Generated by Artificial Intelligence.
The US Copyright Office has overseen the federal copyright system since 1870, with a legal duty to advise Congress and the Federal Judiciary on copyright matters. US copyright law (in common with the UK and most other countries worldwide) automatically protects ‘original works of authorship fixed in any tangible medium of expression, now known or later developed’; and includes literary, dramatic, musical and artistic works.
The guidance clarifies that work containing wholly AI-generated material may not be copyright-protected, if it was not the product of ‘human authorship’. However, the guidance further explains, where a human selects or arranges or modifies AI-generated material in a sufficiently creative way that ‘the resulting work as a whole constitutes an original work of authorship’, then copyright protection may apply.
The US is a world-leader in the development of both AI technology and intellectual property law, and this latest AI copyright guidance is likely to influence the thinking of most other countries that have yet to address AI copyright questions. A handful of other copyright law legislatures have, however, already addressed such questions: Hong Kong, India, New Zealand and the Republic of Ireland, each of which jurisdictions has followed the UK’s pioneering legislative lead.
In 1988, the UK’s Copyright Designs and Patents Act included then unique provisions dealing with four categories of work: ‘in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author shall be taken to be the person by whom the arrangements necessary for the creation of the work are undertaken … the work is generated by computer in circumstances such that there is no human author of the work.’
Accordingly, the author-cum-first copyright owner of a computer-generated artistic work in UK law is the undertaker of ‘the arrangements necessary for the creation of the work’ – this terminology is precisely the same as the Act uses to define a ‘producer’ in the context of determining an author/copyright owner of a film.
Even though these UK provisions prudently anticipated the need to give special copyright protection to computer-generated works, artificial intelligence technology was not as developed in 1988 as it has become in recent times. It is therefore understandable that legitimate questions are now emerging as to whether these provisions are fit for more sophisticated AI purposes today, three and a half decades after their original enactment. And such questioning is of interest and importance not only in the UK and the four other kindred computer-generated copyright legislatures, but also in the US and the wider copyright world that will doubtless be looking for appropriate solutions to AI creativity challenges.
A key question is whether the UK’s special computer-generated copyright provisions are at odds with copyright law’s paramount requirement: that a literary, dramatic, musical or artistic work is the original expression of a human mind. For example, when artists use tools to originate artworks, common-sense dictates that they – not their chosen tools – are the authors. And historically, the notion of authorship came into question with the advent of print, photography and the earliest born-digital imagery.
Furthermore, intellectually and legally, there have been challenges about the authorship of artwork not crafted by ‘the hand of the artist’. Marcel Duchamp and Andy Warhol, for example, both designated rather than personally manufactured in their role as author. The Terms of Service of popular AI Midjourney specify that the user grants Midjourney a copyright licence by using its service; and, subject to this, ‘You own all Assets you create with the Services, to the extent possible under current law … Midjourney makes no representations or warranties with respect to the current law that might apply to you’. Curious terms, which make no reference to the key issue of whether users have, in fact and in law, originated copyright-protectable material.
The ‘human mind’ copyright doctrine features in most intellectual property regimes worldwide, adherence to which may perhaps explain why so many countries have not been attracted to adopting the UK’s arguably questionable non-human approach. In the US, for example, the US Supreme Court ruled as early as 1884 that copyright protection at that time excluded works created by ‘non-humans’ (when dismissing a claim that cameras, not photographers, were image-makers) – a legal precedent evidently influencing the US Copyright Office’s recent guidance on ‘material generated by artificial intelligence’.
EU copyright law adopts the same approach, albeit couched in different language: the expression of an ‘author’s own intellectual creation reflecting his [sic] personality’ is a fundamental requirement for a work’s copyright protection. And so, in Spain, for example, ‘the author of a work is the natural person who creates it’; and in Germany ‘copyright protects the author in his [sic] intellectual and personal relationships with the work’. In Australia, courts have authoritatively declared that works are not covered by copyright if they ‘lack human authorship’.
Guadamuz’s scholarly discourse on this complex subject concludes by referring to the central theme of Dick’s novel: that artificial entities, which are ‘replicants’ of humans, may have no built-in awareness that they are machines and not sentient beings; yet their actions may manifest human traits, which make it difficult or impossible for people to distinguish a human from a replicant.
Artistic works wholly generated by AI tools may proliferate into the future, and continue to pose problems for the world’s copyright law nations – most of which have to date largely rejected, or not yet considered, such works being copyright-protected. Perhaps the UK’s 1988 current legislative approach might offer a widely acceptable way forward, if suitably amended: by disapplying the human originality copyright requirement for computer-generated artistic works – just as it has already enacted in the case of copyright for films.
© Henry Lydiate 2023