ARTIFICIAL INTELLIGENCE, COMPUTER GENERATED WORKS AND DISPERSED AUTHORSHIP: SPECTRES THAT ARE HAUNTING COPYRIGHT

INTRODUCTION

It is common knowledge that the fundamentals of intellectual property law in general and of copyright in particular are constantly challenged by new technologies. However, this assumption is less dull that it seems, if one considers it from a specific angle, that seems to affect the foundational elements of the law. As a matter of fact, there are instances in which the traditional concept of “originality” of works as a threshold for copyright protection appears hard to espouse with the creative process behind certain types of works. I mean to refer here to those cases in which technology is not only a tool for the author or, to use an analogy, a brush in the hands of a painter, but it rather is the author itself. The author is the natural person making the creative choices and as such infusing her own personality into a certain work. This is the settled meaning of “author” or “artist” which traditionally is at the root of copyright law. It therefore was no surprise that at EU level the Infopaq reference case identified the threshold of originality under the Info Soc directive in the expression of the “author’s own intellectual creation”. This definition establishes a necessary bi-univocal relationship between the act of creating a copyrightable work and a human acting as creator, so that there is no originality and hence no copyright where there is no natural person(s) behind a work.

Alongside with the conception of authors as human beings, another pinpoint of copyright as we know it is the identifiability of individuals behind works, as a minimum in the form of the person collating and directing the contributions of multiple authors in the event of works of collective authorship.

Phenomena such as Artificial Intelligence (“AI”) leading to Computer Generated Works (“CGWs”) and the dispersed authorship of works like Wikipedia, with the ensuing fine dust of contributions made by myriads of contributors, are destabilising the copyright construct built around the traditional concept of identifiable “authors” as human beings, to the extent that there are works that appear to deserve consideration for copyright law purposes, while however being “authorless” when looked at from the legacy copyright perspective.

Wikipedia in itself is a function of information technology, because it is premised on the sharing of knowledge of its contributors, made possible by the technicality of it being web-based and, thus, modifiable by anyone, at any time, most often by small fragments of information fed into it by each contributor, virtually simultaneously with the publication of the contents itself.

In the realm of CGWs, notable recent examples of authorless works are provided by the attempts to have computer systems mirror the highest expressions of human arts in history. The Next Rembrandt project (http://www.nextrembrandt.com)  is a remarkable manifestation of the wonders of AI in this respect. The project’s goal in a nutshell was to have a machine produce a brand new Rembrandt painting, as if the Dutch genius himself had painted it. This is signified  by the catching question one can read on the project’s Internet landing page: “Can the great master be brought back to create one more painting?” In order to realise the new Rembrandt piece, computer engineers and art experts teamed up and fed into the system massive volumes of data concerning Rembrandt’s works. As a result, the computer, through a machine learning process, singled out recurrent artistic (both aesthetic and technical) patterns in Rembrandt’s style, which an algorithm elaborated and transformed into a creative output of the system itself, i.e. a new Rembrandt’s painting painted with a 3D printer, the portrait of a Caucasian male, with facial hair, 30-40 years of age, in dark clothing, wearing a white collar and a hat and facing to the right. Who is the author of this work and would this work merit copyright protection? Would it be sufficient, for the attribution of authorship to the researchers, that they mined data from Rembrandt’s real paintings and fed them into the machine?

Deployments of artificial intelligence have however been experimented with respect to virtually any “copyright work”, at least considered in the abstract, within the meaning of Section 1 of the UK Copyright, Designs and Patents Act 1988 (“CDPA”), as eloquently exemplified by Ray Kurzweil’s Cybernetic Poet, which, by using a type of machine learning logic similar to the new Rembrandt’s one, distilled the poetic style of a number of authors and produced its own original poem.

This article briefly discusses the authorship conundrum that from a legal perspective is generated when copyrightable works reflect creative choices attributable to machines. It then moves onto considering another phenomenon that can be equated to that of CGWs, in terms of want of an author: reference is here made to the increased collaborative creativity in the digital space (such as Wikipedia) and the various open content movements (e.g., open source software and open data), where, although the creative output can be entirely ascribed to humans, still each individual contribution is so fine grained, that the result is one of dispersed authorship, if any at all. While in a handful of jurisdiction, such as the UK, CGWs are expressly recognised copyright protection, such protection cannot in general be afforded under copyright law in the majority of jurisdictions. I take the view that this is a suboptimal outcome, which however is not impossible to reconcile with the principles of copyright law, in particular considering certain evolutionary patterns of the same, that seem to permit a refocusing of the rationale of protection from the “author” to the “work” as such and, to this end, calls for reform in the copyright legislation. On the other hand, collaborative creativity and the ensuing dispersed authorship in the web seems to be eliciting a move towards a reshaping of the copyright culture, according to which the ultimate purpose is not the protection of the work from plagiarism as such, but the judicial sanction of the ethics of knowledge sharing.

  1. The historical beacon of copyright law: the author as the creative human being.

The fundamental tenets of global Copyright law[1] historically have been inextricably intertwined with the centrality of the “author”, this typically being the flesh and blood individual behind the creation of a work.

This can be observed in particular if one picks from a few – although fundamental – normative choices of global copyright law. In particular, a selection of such choices, revealing how copyright law has been designed with authors as the central focus, includes the following:

  1. i) the term of protection afforded is as long as the author’s life and shall continue for seventy years after her death;
  2. ii) the concept of “originality” of the work as a prerequisite for protection has been interpreted as an indication that a work to be protected should be the result of the author’s own intellectual creation (or skill and labour in the UK) and, more significantly, that it must bear the imprint of the author’s own personality, showing the author’s exercise of his subjective judgement.

This two elementary norms of copyright law are tributary to what has been defined as the “ideology” of authorship[2], which is rooted in the Romantic idea that invariably associates human genius and authorship. In this context, it is of relevance the historic conception of authorship, which informed the original structure of copyright laws around the globe and which at least formally has not changed: the author is idealized as a creative artist, inspired directly from nature.

This conception seems to imply a hierarchy of artistic productions, where artistically meritorious one would displace, as the only ones deserving copyright protection, those that instead are devoid of any artistic value. However, here a first departure from the original abstraction of the conception of authorship becomes apparent, to the extent that it is a settled concept that copyright law is not concerned with the artistic merit of the work. Anything falling within the categories of copyrightable works may accede to copyright protection[3], regardless of how modest the artistic quality may be. One author, taking stock from this principle of the law as explained in the US Supreme Court’s landmark case Bleinstein v. Donaldson Lithographic Co. [188 U.S. 239], has held that the Bleinstein opinion, with its rejection of the court’s role as adjudicator of the aesthetic value of a work, has left the category of authorship “with little or no meaningful content and none of its traditional associations[4] with its original Romantic idea.

It cannot be given for granted that in Bleinstein the Court did actually play down the role of authorship in copyright, because the Court seemed more concerned with avoiding the risk that requiring a threshold of aesthetic merit to be passed could result in false negatives[5]. On the other hand, the Court itself insisted on the uniqueness that any work may display, which is strictly linked to the personality and singularity of each individual creator, independently from the artistic gradient.

  1. From the law of authors to the law of copyrightable works?

The point that I wish to make here is that authorship in copyright may at a closer look not be an a priori of copyright law, but it is nothing more than it should obviously expected to be, i.e. a legal construct, designed around practical and pragmatic considerations.

It should in this connection be recognized that, as much as the current legal copyright framework seems indeed to gravitate around the “author”, to begin with it is even uncertain who the author is, or, else, what legal qualities a person should satisfy to be recognized as author. This is because neither national, nor international copyright laws provide a clear cut definition of authorship. If one looks at the primary source of international copyright law principles, the Berne Convention, it “provides only limited guidance: while it lists a series of works in article 2 that each Union country is to protect, it does not contain any definition of the term author[6].

National laws, in this context, may provide indirect hints as to how authors should be defined. This is the case, for example, of the UK Copyright, Designs and Patents Act 1988 (“CDPA”), which at Section 9(1) identifies the author in relation to a work as the person who creates it. Similarly, Article 6 of the Italian copyright law (law no. 633/1941 as amended), provides that the original title giving rise to copyright ownership shall be the “creation” of the work, to be intended as the peculiar expression of the intellectual work. However, neither statute defines “creation” and therefore the mind continues to boggle around what it is meant by authorship in a copyright context.

The ensuing uncertainty may appear bizarre, because the first beneficiary of copyright laws is somewhat hidden in the clouds. It is in any event possible to recognize that the concept of authorship, although central to copyright law, lends itself to ambiguity and instability. In my opinion, maybe the most bewildering example of how the concept of authorship in copyright law may be twisted to accommodate the specific policy goals pursued by a legal system is provided by the doctrine of “Work made for hire” under Section 201 (b) of the US Copyright Act 1976. According to this doctrine, the entity (individual or undertaking) who paid to have the work created, whether by an employee or an independent contractor, is considered the author, unless otherwise agreed. Now, it is clear that such a doctrine entirely separates the act of creation from the ensuing legal interest in the work and gives rise to a complete legal artefact, whereby authorship and endowment of the exclusive rights of economic exploitation deriving from copyright ownership are conflated. This doctrine is the sublimation of the vanishing of the author as a real central concept of copyright law. The “author” under the work for hire doctrine becomes so by virtue of a simple contractual covenant. No input into the creative process is required of the “author”, not even that of directing the work of the person hired for the purpose. Hence the concept of authorship as a function of a creative act is traded for the concept of “deemed authorship” justified by the payment of a consideration.

The above is a short tale of a much wider debate whether it is not somewhat of an optical disillusion to keep looking at the “author” as the sole/real source of justification of the global copyright law construct.

Indeed copyright law as such contains clear indications of its lose or at least unstable dependence on the concept of authorship. Contrary to conventional wisdom, at a closer look into the realm of copyright, it may turn out that copyright is not the law of authors, but the law of copyrightable works. In fact, the list of protectable subject matters, with its idea-expression dichotomy, objectifies the (obscure) concept of author, by epitomising the intellectual abstraction into a physical manifestation. The extreme example of “works made for hire” shows that copyright law can do without the author (meaning that the law may tolerate a framework whereby the statutory author is divorced from the actual creator), but cannot do without the “work” or “subject matter” which is the ultimate target and reason for protection.

The question about the role of authorship in copyright is made more pressing by the cases of computer generated works and of collaborative and dispersed creativity of the type that has given rise to Wikipedia or Open Source (“OS”) Software. These two cases, in particular, elicit yet another question, namely whether it is not time to officially recognize authorless productions within copyright laws or even that there may be creativity outside of the copyright world. The two cases are discussed below.

  1. The conundrum of Computer Generated Works: deserving protection but not being protected – the UK exception.

The issue raised by computer generated works (“CGWs”) is two-fold: whether it is desirable to afford protection to such works from a policy perspective and, in the affirmative, how copyright protection of CGWs can be reconciled with the current configuration of the law, which requires the identification of a “human” as the author. Of course this discussion is in particular relevant for those jurisdictions – which are the large majority – that have not already made their choice to expressly acknowledge to regulate CGWs, as is the case for the UK CDPA, which devolves to the specific issue Sections 9(3) concerning authorship and 12(7) concerning duration of the copyright on CGW. The UK case can however be used to enlighten the discussion.

CWGs are those in which the software fashions the final work. The programmer establishes the rules according to which the program functions but the creative outcome is essentially unpredictable. Differently from computer assisted works, in CGWs the creative choices are akin to random or at least unpredictable from a human perspective, as they depend on the machine and not – even partially – on the humans behind it.  One example of CGWs is represented by databases, which  can be automatically arranged. Another notable example has been discussed at the beginning of this article, concerning the new Rembrandt painting created by Artificial Intelligence that elaborated a creative output out of the analysis of large data sets extracted from the meticulous observation of a number of original  Rembrandt’s paintings. The machine learnt from the data and made the artistic choice that can now be admired in the painting, so that, at least phenomenologically, the machine can be said to be the author.

The example of databases is significant in this context, because it demonstrates how under the current EU legal framework copyright protection of databases is conditional upon the existence of a human author. As a matter of fact, under Article 3 of EC Directive 96/9, “databases which by reason of the selection or arrangement of their contents, constitute the author’s own intellectual creation shall be protected as such by copyright”. As observed above, following the EUCJ’s Infopaq’s judgment (Infopaq International v. Danske Dagblades Forening [2009] ECDR 16 Case C-5/08) the “author’s own intellectual creation” is the test currently deployed at EU level to measure originality, as a prerequisite for copyright’s subsistence. Similarly to what has been held under Australian copyright law, the statute in essence demands that that the work originates from a (human) author, to the extent that the «two expressions “author” and “original work” have always been correlative: the one connotes the other»[7]. The characterization of originality/authorship as essentially human attributes leaves no room for the recognition of copyright protection for CGWs, because what the law demands is the exercise of “creative freedom” (see EUCJ in Football Dataco Ltd. v Yahoo Case C-604/10), where the author(s) – intended as human beings – would exercise none in CGWs.

I concur however with those voices holding that the grant of copyright protection to CGWs accords with copyright policy goals. In particular, the emphasis here should be placed on the desirability of protecting a work, although –generated by the artificial intelligence inside a computer, which more often than not will be unique, the resultant of considerable technological effort and investment, which would increment the existing stock of knowledge and innovation. As has been observed, “given the low standard of intellectual effort required to confer originality, copyright policy does not so much encourage great mental exertion but its productive result – a new and potentially valuable work[8]. The ensuing prohibition on copying would be socially beneficial, because it would secure that something new is created. On the other hand, the importance that the digital economy has gained nowadays makes it irrational to blatantly exclude from copyright protection anything qualifying as a CGW, as this might work as a disincentive for heavily computer dependent creative businesses.

The case for protection of CGWs would confirm that the concept of authorship can be disposed of vis-à-vis the concept of “work” deserving protection, and replaced by a concept of “deemed authorship” and by a substitute for originality. The existing legal framework, however, calls for a legislative reform, absent which CGWs would remain unprotected. Works like the new Rembrandt would remain unprotected in the vast majority of jurisdictions, to the extent that there is no specific provision covering CGWs and the legal framework is premised on principles that are in clash with the possibility of affording protection where no human input in the ultimate creative choice can be spotted. The UK can show the way forward in this connection. The UK  seems in fact to have veered towards the solution that grounds protection in the concept of deemed authorship, whereby the author of CGWs shall be taken to be the person “making the necessary arrangements” for the creation of the work and such arrangements are likely sufficient in the UK to satisfy the “labour and skill” test of originality. At the same time the provision of Section 12(7) of the CDPA on duration might lead to the qualification of CGWs as entrepreneurial works and the relevant right akin to a related right, rather than full-fledged copyright. Another, maybe more acceptable solution at EU level, could be to adopt a similar approach to that applied for the protection of databases in the EU under the sui-generis right, where the protection trigger is the substantiality of any investment in the making of the database. This solution could overcome the hurdle posed by Infopaq that sees originality in one’s own intellectual creation.

  1. The case of collective intelligence of Wikipedia and Open Source projects: yet another blow to the centrality of authorship.

The traditional concept of “authorship” is no les shaken when looking at creative products such as Wikipedia or OS software.

The issue is in the first place one of scale and of clash between positive copyright law and the norms and expectations of the communities that contribute to the making of the pages of Wikipedia or of the lines of code of OS software. The phenomenon we are faced with is in fact one of mass collaboration of countless individuals that cooperate inspired solely by their will to share their creative inputs on a copyright free basis.

This configuration is at odds with the straitjacket of copyright law, which is a default rule requiring the identification of the author(s) and automatically vesting in them certain exclusive rights of ownership. Under the traditional copyright framework, Wikipedia (if not as a whole its pages) qualifies as a work of joint authorship, pursuant to Section 10(1) CDPA. However, as each page is constantly changed, ordinarily by small increments by a myriad of contributors, as has been observed[9] it is difficult to discern which (if any) of the contributors have made the right kind of contribution, as required by the case law, in order to identify the “collective author” under Section 10(3) CDPA. A related issue concerns the identification of the contours of the work and of the relevant originality requirement: on the one hand, the work is constantly in progress, so as to raise the question whether copyright law as it stands is compatible with a work which is not intended to be finished; on the other hand, the increments added to each pre-existing page (or line of code in the OS example) by the contributors might be seen as derivative works. However the often minimal size of each increment puts into question the actual “materiality of the change” that would be required in order for a derivative work to meet the originality requirement.

It is apparent from the above that the ethics of sharing of knowledge that are at the root of phenomena like Wikipedia have little in common with the consolidated norms of copyright. Indeed, the members of the community have no interest in ownership claim and are solely animated by the desire to create a copyright free area, where creativity can be unleashed. This phenomenon of “social production”[10] also detracts from the traditional justification of copyright law as an incentive to creativity. At this junction comes the consideration of the copyleft licensing, as a model that the Wikipedia community (similarly to the GPL of the OS movement) has adopted as the default regime, which can be summarised as “some rights reserved” as opposed to the “all rights reserved” default rule of the statutory copyright regime.

Wikipedians, in fact, are bound to release their contributions under Creative Commons Attribution-Sharealike (CC-BY-SA), which in its very essence allows anyone to change and use, including for commercial purposes, the work so released, provided that the original authors are credited and that any later versions of the work are further released under the same licensing scheme. This is usually referred to as the viral characterization of copyleft licenses.

Copyleft and open source licenses have to come to terms with the apparent contradiction of having to rely on copyright, while attempting to abolish it. However, this is because the relevant communities have no choice other than playing with the rules set by the applicable legal regime: hence, the easiest solution is for the authors to relinquish copyright by way of a private declaration/contract, thereby donating the result of their intellectual efforts, while virally forcing those wishing to participate in the creative community to adopt the same copy-free approach.

The example of social production vividly shows how the current global copyright regime is unfit for new forms of mass creativity, which obfuscate several pillars of traditional copyright law, beginning with the very concept of authorship. As it has been argued, multi-author mass collaboration with no hierarchical organization is a game changer, and the ensuing creativity is likely not authorship[11] and does not even need the recognition of the copyright armour, as the norms of the community that nourish the sharing of knowledge have already found their way into a copyright free world.

CONCLUSION

CGWs and (Wikipedia-like) collective works of dispersed authorship are a tale of inadequacy of settled principles of global copyright-law. At the risk of oversimplification, it could be said that too little copyright law in the case of  CGWs is counteracted by not the right type of protection in the case of collective works. The two cases share their common denominator in the lack of an author (in the traditional sense) and call for the rethinking of established principles of copyright. The law has developed over time around the justification of copyright protection in exchange for the contribution made by other humans to the creative acquis. A the same time, the entry level of creativity required for the recognition of copyright protection has been set very low, so that anything in which the human genius can be spotted is protected. CGWs reverse the creator-work relationship, in one in which, at one extreme, highly artistically meritorious works are generated by AI, which in its turn may be able to process input information much more accurately and in larger volumes than humans. On the other hand, the dispersed authorship of Wikipedia-like works and the ethics of sharing behind them warrant for the law to change paradigm and acknowledge the choice of the community not to allow the privatization of knowledge not as a contractual right but as a real intellectual property right enforceable erga omnes.

[1] Of course there is no such thing as a “global” copyright law. However, a lower common denominator shaping copyright law throughout most of the world countries is provided by international treaties seeking to level certain principles of copyright, such as the Berne Convention Berne Convention for the Protection of Literary and Artistic Works, the TRIPs Agreement or, at regional level, by the instruments adopted by the European Union, which has to a certain extent harmonized copyright law.

[2] Peter Jaszi, Toward a Theory of Copyright: the Metamorphoses of “Authorship”, Duke Law Journal, Vol. 1991, No. 2 (Apr. 1991), pp. 455-502.

[3] There are exceptions to this rule, a notable one being enshrined in Article 1, no. 10), of the Italian copyright law, which requires that industrial designs, in order to be protected by copyright law, must present “artistic value”.

[4] Peter Jaszi, cit. at fn. no. 3, page 483.

[5] The Court states: “It would be a dangerous undertaking for a person… to constitute themselves final judges of the worth… At one extreme some works of genius would be sure to miss appreciation. Their very novelty would make them repulsive until the public has learned the new language in which their author spoke”.

[6] Sam Rickeston, The Berne Convention 1886-1986, chapter 6.4 (1984)

[7] J. McCutcheon, The Vanishing Author in Computer Generated Works: A Critical Analysis of Recent Austrialian Case Law, 36 Melb. U. L. Rev. 915 2012-2013, at page 935 with reference to case-law.

[8] J. McCutcehon, cit. fn. 7, at page 954

[9] D. Simone, Copyright or Copyleft? Wikipedia as a Turning Point for Authorship, in SSRN, at page 17.

[10] So termed by Elkin-Koren, Tailoring Copyright in Social Production, [2011] 12 Theoretical Inquiries in Law 309

[11] C. Sunstein, Infotopia: how many minds produce knowledge, (OUP 2006) pg. 153