The institute of the ‘Fictio Iuris’ is deeply rooted in the western legal tradition. It was created two millennia ago, during the times of the Roman Republic. Its definition is: “When the Law, for a just cause, artificially recognize the existence of something that is not really existent or consider as happened something that has not really happened ”. The most cited example of the fictio iuris in roman times was the Lex Cornelia, where if a soldier dies in captivity, it was considered dead from the moment of his capture. This was due to the fact that a prisoner of war was lawfully stripped of all of his rights, but this caused the unjust nullification of his testament. Therefore, the ratio of the fictio iuris was to solve a situation which was seen unjust without redrafting an entire set of laws.
The institute of the fictio iuris remains active even in the modern times. Sometimes an institute derived from the fictio iuris from the Roman epoch is present even today, like the institute of adoption. Adoption results in a person being considered a son or a daughter of someone even if there is no real biological connection between them. Another example is the institute of tacit consent, which is widely used. Tacit consent is used especially in administrative law, to have a more efficient and less costly public administration. It is used also in private contract law. Tacit consent is used in contract law for example when one party, by not explicitly express its consent or dissent to renewal or dismissal of a contract, cause the clause of tacit consent to activate its predetermined effects. The last and most important example is the institution of the Corporate Personhood. The institution is essential for today’s society and without it many actions that benefits the economy could not be performed. It is clear that a corporation is not a human who is capable by itself of performing acts that have legal repercussions, yet we accept this legal fiction in our daily lives without much thought .
The possibility of AI creating works of authorship was granted through a mix of great advancements in the last twenty years: from the big data revolution to the constant increase of computing capacity and diminishing cost as described in Moore’s law, from the refinement of machine learning techniques to the development of neural networks.
AI is now more than ever capable of creating copyrightable material, but the problem is that Copyright laws all around the world require a human to be behind the Copyrightable creation.
This article will in the first part establish some common grounds to better understand the capabilities of AI in the field of art and copyrightable materials in general. Then, it will delve into why in the actual copyright legislative system AI’s artistic capabilities cannot entirely flourish. Finally, we will see why and how the fictio iuris can help us fix the issue and why it is only a temporary solution.
The current State of the Art
The most important pieces to understand the current situation are ‘machine learning’ and ‘deep learning’. Machine learning (or ML for the rest of this article) is «a type of artificial intelligence in which computers use huge amounts of data to learn how to do tasks rather than being programmed to do them ». Deep Learning (or DL for the rest of this article) is «a type of artificial intelligence that uses algorithms (= sets of mathematical instructions or rules) based on the way the human brain operates ». Both these techniques, in conjunction withthe huge amount of data available thanks to the Big Data Revolution and the evermore expanding capacity of calculation by modern computers, give AI’s the possibility of learning how to create copyrightable artworks. AI can therefore be fed ten of thousands of books, or paintings, or sculptures and learn through ML or DL to then create something new. These AI are different from the ones who only make collages of previously fed creations.
In the past, the computer was just seen as a tool of the human artist. Hundreds of thousands of books are written every year using a computer. Painting with the use of a computer is common since the early years of informatics, like the last artworks of Warhol before his death . Now with all the advancements available, computers are not only tools but artists themselves.
In the field of music, for example, in 2019 a neural network (DL) was made to analyze all the previous songs presented at the Eurovision contest. Then, the neural network wrote and compose a song resulting in the one titled ‘Blue Jeans and Bloody Tears’ which was presented at the 2019 Eurovision contest .
Regarding paintings, now AI’s works can be purchased even in the most important auction houses. For example, in 2018 a painting made by an AI was sold at Christie’s for $432.500 . The AI was managed by a Parisian collective which used a technique called Generative Adversarial Network (ML). The AI was fed thousands of portraits from the 14thto the 20thcentury and the results was displayed and sold at the auction house. There was also what we could consider a provocation, since the traditional human’s signature on the painting was substituted with the equation used in creating the painting itself, a sort of AI’s signature.
While we are still far from having an AI having its book reaching the top of the monthly selling charts, AI’s generated books are starting to be sold. For example, in 2018 a book was marketed as the first entirely written by an AI. The book is titled ‘The Road’. A particular case indeed, as the data input for this AI was not only made with already existing books, but also with all sorts of outside feeders like a GPS, a camera and a microphone. All these alternative data input feeders were mounted on a car which roamed the United States, like an author searching for inspiration. The result was ‘1 the Road’ . The novel is now available to buy in both online and offline.
The examples could be many more, yet we need to continue the discussion. We need to ask ourselves: what about the copyright on these incredible feats of technology and inventiveness?
The issue of AI’s creations copyright
Issues in AI’s creation copyright are not new. Some nations have already tried to solve the issue, most if not all of them are part of the Commonwealth, like Ireland, India, New Zealand and Hong Kong. In particular, in the United Kingdom,section 9.3 of the Copyright, Designs and Patents Act (CDPA) of 1988 states: «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». Section 178 of the CDPA defines AI’s generated works as: «generated by computer in circumstances such that there is no human author of the work». While it is an important step forward, its still too generic and has various problematic sides. What if someone buy an AI and then just push the button to create a painting? What is really the definition of ‘arrangement necessary for the creation of the work’? What if an AI is created and let loose to decide what it does, and in the end creates a copyrightable work? The cost of going to court due to an unclear law is in my opinion still outweighs to profitability of developing an AI.
Other intellectual property law systems can be seen as hostile to AI’s only creations. For example, in the European Union, its member states’ legislation on copyright is human-centric . We cannot obviously blame anyone, as during the long history of copyright it is the first time that something else than a human is capable of creating something that could be included in copyright . The Berne Convention and all the other pieces of legislations, conventions and charter are structurally made and developed with a human as the only possible author. The safeguards provided by copyright itself are constructed as a way to compensate and sustain the author in its creative activity. The term creation is also conceived having in mind the ‘spark of creativity’ residing in the human brain. The same could be said regarding the copyright system of the United States. While there are few issues concerning conceding copyright to work created by AI in cooperation with humans, where the copyright is obviously given to the human, a judge will have a lot of difficulties in cutting the gordian knot of AI’s autonomous creations.
Therefore, the main issue is that we have something other than a human creating a work which in theory, if it was made by a human, could be copyrightable. As it is always been, a potential market and field of technique which is ignored or even obstructed could be stifled and not express its full potential.
Why use the fictio iuris for resolving the issue
One of the most used justifications for the establishment of Copyright as it is written in constitutions and laws around the world is to further progress and cultural enrichment of society and its denizens. Another widely used justification is to remunerate those who take on the task of creation with rights to control for a limited amount of time the circulation of those works. These justifications are in general used for all rights incorporated into the institution of Intellectual Property. AI’s perspective could be an incredible enrichment for Society as a whole, a point of view never seen before, capable of making connections in ways we could never imagine. Thus, we should create a favorable environment in which these new incorporeal artists could express themselves. To reach this objective we should create a favorable legal environment, one in which the uncertainties of an unprepared and old legislation are resolved. The time for legislators to act is running out, and this temporary solution could give them more time. Time to assess the situation, time to gather enough expertise to then draft better laws to resolve the conundrum of AI’s authorship.
Therefore, we have a potential increment for our society, a legislative landscape which does not favor it and the possibility of using the legal fiction to temporarily solve this legal conundrum without needing to rush legislators to solve it. That is why, in my opinion, the fictio iuris could be a useful institute to apply in this case without having to betray to nature of the institute itself.
How to use the fictio iuris to solve the issue
While the general solution is the one cited in the title, its details will differ from country to country, with the unified objective of granting legal certainty to the authorship of AI’s only creations and the possibility of being recognized anywhere in the world. The fictio iuris itself must be nonetheless coherent, without an excessive dispersion from the original intentions. The intention here is to have a fictious contractual relation which would permit the transfer, from a fictitious person (the AI) to a real physical person, of the copyright in its entirety or at least to the extent that various national and supranational rules permit. The relation will therefore be between the AI and an explicitly nominated physical person. In most of the cases an AI capable of creating copyrightable material is part of an already established company, therefore the physical person will be part of the company’s workforce. The obvious consequence will be that the designated physical person will be selected by the company. This would mean that the selected physical person will have previously signed all the necessaries contractual clauses in its employment contract to subsequently pass over the Copyright of the AI generated works. But, even if it is only an artist or a programmer creating an artistic AI, he or she will be capable of having the copyright protected with the same possibilities as the ones granted to corporations.
We will now look at how to apply this legal fiction in the United States and in the European Union.
Starting from the United States, copyright could be easily transferred through the designation ‘work for hire’ on AI’s autonomously generated work. With the ‘work for hire’ copyright is transferred at the moment of the creation of the work from on party to the other. This can happen in two situations: when it is created by an employee as part of his or her contractual duties and when a certain type of work results from an express written agreement between the creator and a party specially ordering or commissioning it . We already have a legal fiction, at least the if we look at this transfer with European eyes, since the party to whom the copyright has been transferred to is considered the author. Thus, in the American legal system, it could be easy to extend this fiction to the contract, in which one of the parties is considered human and capable of conducting legal actions like negotiating and signing a contract. It could also develop into a paradox, where a fictitious person (the AI) transfers its copyrights to, in the end, another fictitious person (the corporation).
In Europe the situation is more complicated. Differently from the United States, in Europe there are certain rights, the moral ones, which cannot really be alienated from the original author. Thanks to this influence, moral rights are difficult if not impossible to alienate. A solution to accommodate these difficulties without needing to divert too much from the main objective of the legal fiction, the AI could be considered an anonymous author. In this case the human, to whom the copyright will be transferred to, could be considered as the agent of the AI with all the already established solutions available in every country with a modern copyright system. Anonymous authors with a pseudonym could also work, as this would maintain the human as the ‘center of the legal focus’ requiring less fictitious action to bring about the intended result we are trying to reach.
Another point to ponder on is the duration of this copyright. In the United States, a work for hire will result in copyright remaining valid for 95 years if published or 120 years from the creation of the copyrighted work, whatever expires first. In Europe, since the most viable solution is the legal fiction of anonymous or pseudo anonymous content, the solution is also at hand as in most member States of the European Union and other European States regulate the subject matter harmoniously. The common duration is 70 years from the publication or creation, whichever comes first.
However, these solutions could be put in crisis since the algorithm does not have a stable identity. In reality, even putting a minimum of agency on something so ethereal as code opens the door to a myriad of problems. How much of an algorithm can change before it cannot be identified as the same? Just one character or more lines of code? Only if we suspend these questions, we can put in motion the afore mentioned legal fiction.
To simplify this, we could apply a solution which uses legal fiction and an already establish solution which is akin to what has been shown before. A solution which could be applied for both the American and the European framework of laws. In fact, we could consider the legal relation between an algorithm and the human to whom to right is transferred to as an engagement between an anonymous artist and a publisher. This solution is widely accepted throughout most countries with a copyright system in place. In this case there is no real need of even creating a fictitious party, instead most if not all the rights related to the content are on the publisher’s shoulders and also in most countries duration is connected in this case to the time of the publication of the copyrighted material. Hence, there would be no uncertainty and neither there would be problems in the duration of the copyright since the usual duration is 70 years for works of anonymous authors that are being publicized. The legal fiction is then that behind the wall of anonymity there was a human and not an AI.
So, we could in full respect the Berne Convention for the Protection of Literary and Artistic Works and taking its article 7.3 establish a copyright duration of 50 years from the moment of its publishing. Also, to take out any uncertainty, the publisher of the AI autonomously generated work will be considered responsible for exercising the rights and to be responsible of the wrongs of the work itself. Thus, this solution can achieve certainty and a reasonable timespan to remunerate whoever created or purchased the algorithm which created the work itself.
Conclusion on a temporary solution
The traditional institute of copyright can only temporarily contain AI’s autonomously created works. The weight of the amount of potential material is too much to accommodate. It will come a time where we will have to decide between favor or obstruct AI in its path to artistic creation. Long term solutions are starting to pop up, but they are still to be developed into fully fledged ones. In the meantime, this is a proposal that could give more time when it is running out. Legal systems around the world are already full of legal fictions, as some adjustments or situation are better to be nudged into an already existing legal category, temporarily or even forever, like adoptions or the legal capacity of corporations.
I believe that by favoring AI in the artistic field, we will gain access to a whole new universe of concepts, stories and music. What was previously hidden behind the limits of the human intellect could now be unlocked by cultivating the new possibilities given by the continuous development of new, more potent Artificial Intelligences.
Personal translation from the original «dicesi quando la Legge per giusta finge essere o non essere stata qualche cosa, che non è e non fu, ma può essere o essere stata».Pothier R. G. (1842) Le Pandette di Giustiniano, Volume IV, Tipografia Giustinianea, Venezia, p. 502
At the legal standpoint, many important and interesting discussions have been constructed upon the institute of the corporate personhood and its responsibilities, and this legal fiction is used since the Medieval times. To know more see Blair M. M. (2013) Corporate Personhood and the Corporate Persona, in Illinois University Law Review, pp. 785 – 820; also see Lyman J. (2011) Law and Legal Theory in the History of Corporate Responsibility: Corporate Personhood, in Seattle University Law Review, Vol. 35, pp. 1135 – 1164
From the Oxford Dictionary, available at https://www.oxfordlearnersdictionaries.com/definition/english/machine-learning (Accessed on 14/04/20)
From the Oxford Dictionary, available at https://dictionary.cambridge.org/dictionary/english/deep-learning (Accessed on 14/04/20)
Stinson L. (2014) An Amazing Discovery: Andy Warhol’s Groundbreaking Computer Art, in Wired.com, Available at https://www.wired.com/2014/04/an-amazing-discovery-andy-warhols-seminal-computer-art/ (Accessed on 14/04/20)
More information available at https://www.nimshap.com/ai_eurovision (Accessed on 05/04/20)
Quackenbush C. (2018) A Painting Made by Artificial Intelligence Has Been Sold at Auction for $ 432.500, in Times.com, Available at https://time.com/5435683/artificial-intelligence-painting-christies/ (Accessed on 05/04/20)
Merchant B. (2018) When AI Goes Full Jack Kerouac, in Theatlantic.com, available at https://www.theatlantic.com/technology/archive/2018/10/automated-on-the-road/571345/ (Accessed on 15/05/20)
CF Di Fiore R. (2020) Magic of art created by computers, in Medialaws.eu, available at https://www.medialaws.eu/magic-of-art-created-by-computers/ (Accessed on 16/04/20)
CF Ben-Tal O. et al (2019) Artificial Intelligence and Music: Open Questions of Copyright Law and Engineering Praxis, in Arts, Special Issue Music and the Machine: Contemporary Music Production, Vol. 8, Issue 3, pp. 115-130, Available at https://www.mdpi.com/2076-0752/8/3/115, DOI: https://doi.org/10.3390/arts8030115
There have been in the past controversies about paintings or photographs made by animals, but this happened due to fortuitous events or by training an animal in doing something its mind cannot even conceive as a work of art. We are here discussing about an activity and its results that are systemic and that could be a stable, everyday occurrence in the future
United States Copyright Office (USCO), (2017 revision) Circular 30, pp. 7, Available at https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=12&ved=2ahUKEwiRiOPXmf7oAhXVtHEKHaHECWcQFjALegQIARAB&url=https%3A%2F%2Fwww.copyright.gov%2Fcircs%2Fcirc30.pdf&usg=AOvVaw1oDWMGBPcLWvnm70w_bnvt (Last seen 16/04/20)
«In the case of anonymous or pseudonymous works, the term of protection granted by this Convention shall expire fifty years after the work has been lawfully made available to the public. However, when the pseudonym adopted by the author leaves no doubt as to his identity, the term of protection shall be that provided in paragraph (1). If the author of an anonymous or pseudonymous work discloses his identity during the above-mentioned period, the term of protection applicable shall be that provided in paragraph (1). The countries of the Union shall not be required to protect anonymous or pseudonymous works in respect of which it is reasonable to presume that their author has been dead for fifty years».