Artificial intelligence - the new author.
1.Introduction.
"Artificial intelligence" (or "AI") as a simulation of human knowledge by self-learning machines[1] (computer systems) is increasingly entering the field of copyright, and for nearly 8 years has been "creating" in various fields of art. In 2016, a group of museums and researchers in the Netherlands presented a work of fine art in the form of a portrait titled "The Next Rembrandt" - drawn and generated by a computer that has analysed hundreds of works and thousands of fragments of them by the 17th century Dutch artist Rembrandt Harmsenchon van Rijn. The interesting thing about this is that "The Next Rembrandt" is a computer-generated three-dimensional painting developed by a face recognition algorithm (the so-called "face recognition protocol") that scanned data from 346 known paintings by the Dutch artist in a technological process that lasted 18 months. The portrait consists of 148 million pixels and is based on 168 263 fragments of Rembrandt's works stored in a specially created and compiled author database. The project is sponsored by the Dutch banking group ING in collaboration with Microsoft, the marketing consultancy J. Walter Thompson and consultants from Delft University of Technology, Mauritzhaus and the Rembrandt House Museum.
At the same time, a short novel written by a Japanese computer program (again in 2016) reached the second round of the national Hoshi Shinichi Literary Award. The novel the artificial intelligence authored is called "The Day A Computer Writes A Novel". In previous years, the Hoshi Shinichi Contest has also been open to non-human applicants, but in 2016, for the first time, the award committee received a proposal from an artificial intelligence. Of the 1,450 entries submitted to the competition, 11 were partially written by a computer program.
Across the Pacific, Google-owned artificial intelligence company Deep Mind created software in 2018 that can generate music by listening to recordings. Post-facto, Google's artificial intelligence researchers unveiled an even newer muzak project, under the name "MusicLM." This is an artificial intelligence model that can generate music with high accuracy from text. "MusicLM" creates music at a constant frequency of 24 kHz over several minutes, modelling the conditional process of sound generation as a hierarchical sequence-to-sequence modelling problem. According to the research paper, "MusicLM" trained artificial intelligence on a dataset of 280,000 hours of music, with the goal of creating songs that have artistic meaning and message. The "MusicLM"[2] researchers also claim that their model outperforms previous AI systems(including "Deep Mind") in terms of both sound quality and adherence to lyrics.
In fact, "robotic artists" have long been involved in various kinds of creative activities, with computers creating crude works of art since the late 1970s. Until recently, most of these computer-generated artworks relied heavily on the creative input of the programmer, but today we are in the focus of a technological existence that validly realizes computers as de facto authors in the creative process. These new technological realities directly relevant to copyright doctrine are due to the daily development of so-called "machine learning" software, a subspecies of artificial intelligence that creates autonomous systems capable of learning without being specifically programmed by a human.
Hence the question arises for every lawyer involved in copyright law - whether artificial intelligence can be qualified as an "author", analogous to any natural person (human being) and whether it possesses and/or can be attributed an authorial contribution in the creation of a specific work, which can be qualified doctrinally and factually as "artistic" or "scientific" and in this context be attributed the role of a fit subject of copyright, within the meaning of Article 3 of the Copyright and Related Rights Act? This paper aims to answer this important theoretical question - to qualify but also to explain the role of artificial intelligence in copyright law as a legal, scientific and creative phenomenon both today and in the future.
2.Technological and substantive aspects of artificial intelligence from a copyright perspective.
A computer program developed for the purpose of machine learning has a built-in algorithm that allows it to learn from the input data, evolve and make future decisions that can be directed or independent. When applied to art, music, and literary works, machine learning algorithms actually learn from input provided by programmers. Thus, the corresponding computer system(artificial intelligence) arranges this data logically and systematically to generate a new work of science and art, making independent decisions throughout the process to determine what the newly created artwork should look like. An important characteristic of this kind of artificial intelligence is that although programmers initially set specific parameters for the machine (computer), the artwork is actually generated by its software itself, called a "neural network," in a process similar to the thought processes of humans. All of the above has different manifestations from a business, financial and legal point of view, which should be highlighted for reasons of correctness.
2.1. Commercial aspects of the problem.
From a business point of view, it should be said that the way in which the laws of different territories treat new types of creativity produced by machines may have particular market specificities. Artificial intelligence is making dynamic inroads into the creation of works in the fields of music, journalism, games and fine art, animation, the comics industry, etc. Worldwide legislative practice and doctrine stands firmly(at least for now) behind the view that works created by artificial intelligence cannot be subject to copyright, as they are not subject to human artistic and scientific potential. In this sense they can be freely used and licensed to any legal entity. This is naturally bad news for the companies behind the development of an AI project and the artistic works generated by it, which also pursue a certain commercialisation. Imagine that a multimillion-dollar investment is made in a medical research system, or a program generates music for video games, movies, or creates an image of comic book characters that are sold online as a series. The motive that such an AI project does not fall under copyright regulation and anyone can use its achievements for personal purposes and/or for the development of their own advertising (for example) project borders on the lyre of legal logic justifying the huge investments in the development of technology and artificial intelligence.
While it is difficult to determine the exact business impact that the aforementioned problem is having on the creative industry, it should definitely be noted that all of this may have a deterrent effect on investment in automated, self-learning computer systems in the future. Here again, the question already asked arises - if developers of AI algorithms are certain that works created through artificial intelligence do not meet the legal framework qualifying them as possible objects of copyright, then what is the incentive to invest in such developments? On the other hand, the use of artificial intelligence to tackle highly labour intensive problems - for example in the field of scientific research and the creation of an encyclopaedic knowledge database and/or information fusion for a specific purpose(for example, the creation of an innovative drug in medicine and pharmaceuticals) cannot be qualified (in my opinion) as a lack of "creative contribution", since tackling such a problem of heavy scientific analysis by a self-learning machine would sometimes take days(given the exact parameters methodology and subject matter for the The most direct example of this is databases - are they a copyrightable asset? The affirmative answer to this query is more than obvious. Hence - who would more quickly and efficiently organize, assemble, and synthesize a detailed, encyclopedic database on a scientific topic - man or machine? The answer I find again is clear. Does artificial intelligence then have a copyrightable contribution, albeit as "not human"? I hope in the future that answer would be - "yes".
3.Substantive aspects of artificial intelligence.
I would like to begin this part of the article by pointing out that, to my knowledge, the granting of copyright in works created by artificial intelligence has never been explicitly prohibited. Hence, I find illogical the legal opinions that are based on the extremely limited view that "if a work is not created by a human, it cannot be subject to copyright".
3.1. Views of Anglo-Saxon jurisprudence and doctrine.
The domestic practice of the so-called U.S. Copyright Office, administered by the Library of Congress in Washington, D.C., is to "register an original work of authorship only if the work is created by a human being." This position stems from U.S. case law[3] which specifies that copyright law protects only "the fruits of intellectual labor" that are "based on the creative faculties of the mind."
Digital art, poems, and books created using tools such as "DALL-E," "Stable Diffusion," "Midjourney," "ChatGPT," or even the recently released "GPT-4" would not be protected by copyright if they were created by people using only a textual description or prompt to the machine to create a post facto derivative work, the director of the U. S Copyright office" - Shira Perlmutter.
"If the traditional elements of authorship of a work are created by a machine, the work lacks human authorship and the office will not register it," she wrote in a document outlining copyright guidelines on the topic of artificial intelligence. For example, when AI technology only receives a prompt from a human and creates complex written, visual or musical works in response, "the 'traditional elements of authorship' are defined and performed by the technology - not the human user," Perlmutter argues.
Another example from the United States is the fact that "an image generated by artificial intelligence does not have the 'human authorship' necessary for protection" - i.e., in the event that a work of art is created entirely by software, there is no copyright protection for it. On this topic, the Zarya of the Dawn case is of particular interest.
In September 2022, the U.S. Copyright Office granted the first-ever registration of a comic book created using artificial intelligence, a fact that was quite surprising and revolutionary in itself, given the debated opinion of its director. Christina Kashtanova, the artist who created "Zarya of the Dawn", explained that the department asked her to provide details of the process proving that there was a human involvement in the creation of this graphic novel. Kashtanova sent information that a large part of the comic was created using artificial intelligence, on the basis of which the US Copyright Office contacted the artist, informing her that they were starting proceedings to revoke their previous decision to grant copyright in a comic that was created using artificial intelligence. The rationale for this decision is that copyrighted works must be created by humans in order to receive formal copyright protection[4].
The author has been granted 30 days to appeal the decision of the US "Copyright Office", during which time copyright protection on the commented work is still valid. In the meantime, her attorney submits an opinion in which he develops detailed reasoning that human participation in the creation of art with artificial intelligence can be considered copyrightable subject matter because the creative process through generative artificial intelligence is "substantially similar to the artistic process of subject selection, time of day, angle, and framing of the image by the photographer," etc.
At present, the "Zarya of the Dawn" case is pending in federal court in Washington, D.C., in the hope that it may yet bring objectivity and clarity to the right of artistic works generated by artificial intelligence to receive copyright protection in the future, with the goal that this motive may serve as a precedent in American copyright doctrine and practice that breaks with the current, in my view incorrect, view of the copyrightable contributions of artificial intelligence.
The second option, where authorship is vested in the programmer, is found in several countries such as Hong Kong, India, Ireland, New Zealand and the United Kingdom and Australia. On the subject, the Australian Court of Appeal has enshrined the view that "a work created with the intervention of a computer cannot be protected by copyright because it is not created by a human."[5]
This approach is best advocated in UK copyright law, section 9(3) of the 'Copyright, Designs and Patents Act' or CDPA, which states that 'in the case of a literary, dramatic, musical or artistic work which has been created by means of a computer, the person who has taken the steps necessary to create the work shall be deemed to be the author. " The same law defines a computer-generated work as a work that "is generated by a computer under such circumstances that there is no person who is the author of the work." The idea of such a provision is to create an exception to any "human authorship" requirement by recognizing the labor that goes into creating a program capable of generating works, even if the creative spark is taken by the machine. I find this vision a step in the right direction.
3.2 European practice and doctrine.
Recently, the Court of Justice of the European Union (CJEU) motivated its landmark decision in the Infopaq case[6] with the view that copyright applies only to original works and that originality must reflect "the author's own intellectual creativity". An expansive interpretation of this decision leads to the conclusion that an original work of art must reflect the handwriting and personality of the author, which reasoning gives us the logic that a human author is necessary for the existence of a work of authorship.
Although European Union ("EU") copyright legislation consists of eleven directives and two regulations harmonizing the fundamental rights of authors, performers, producers, and broadcasters, copyright is still primarily governed by the national laws of each EU member state, particularly with respect to the rules for recognizing authorship and allocating its economic and moral powers.
The basic understanding of EU copyright law is that, in order to be protected by copyright, two conditions must be met: the creation must be an "artistic or scientific work" and the original author of said work must be the person who originally created the work or acquired the copyright by express stipulation to that effect.
The EU acquis explicitly harmonises the following specific categories of subject matter protected by copyright: computer programs, databases, photographs and possibly works of fine art. These works are protected if they are "original in the sense of being the author's own intellectual creation". Like the Berne Convention, the EU copyright acquis is primarily based on the tradition of copyright (droit d'auteur): copyright protects original expression originating directly from the human creator. The Berne Convention does not define "author" of a work, leaving this to the contracting parties, but its text and historical context strongly suggest that "author" and "authorship" for the purposes of the Convention refer to the natural person who created the work.
All of this contradicts the view that copyright protection is generally afforded to human authors. It is fair to note that the Convention's provision on moral rights (Article 6bis), which is expressly granted to "authors", emphasizes that its minimum standards for copyright protection are triggered only by acts of human creativity. But are there exceptions to this view imposed by the new technological realities of the 21st century?
On the one hand, the CJEU case law on "originality" as the main criterion for "copyright work" is entirely based on the notion of a human being who performs creative acts reflecting "creative choices". The CJEU justified this view in Painer[7], a case concerning the creation of photographs by artificial intelligence, on the ground that 'by making different choices, the author of a portrait photograph can give the work created his "personal imprint" (as opposed to those created by AI). In another landmark case, Infopaq,[8] involving the defense of newspaper articles against unauthorized scanning, the Court focused on examining the originality of the "form, manner of presentation of the subject matter, and linguistic expression," explaining that for literary works, the author's "free and creative choice" refers to the selection, sequence, and combination of words, although recognizing that words alone do not constitute an intellectual creation. The ECJ therefore adds that it is only through 'the choice, sequence and combination of these words that the author can express his creativity in an original way and achieve a result which is an intellectual creation'. All of the above suggests that even a combination of fairly obvious choices in the design, implementation and editing of a piece of information, data, colours and images, aided by artificial intelligence, may be sufficient to create a new work of authorship.
This is why in 2020 the European Commission proposed a 'four-step test' for the presence of 'AI technology authorship', concerning a report on 'Trends and developments in Artificial Intelligence - Challenges for the intellectual property rights framework'. Such a test provides the following interrelated criteria that must be met for a work created by AI to qualify as a "work":
Step 1 - The work must be one in the field of literature, science or art;
Step 2 - The work is the product of human intellectual effort;
Step 3 - the work is a function of originality/creativity (creative choice);
Step 4 - the work is a type or mode of expression.
The European Commission's study on the issue of "works created by artificial intelligence" has reached several conclusions and recommendations on EU copyright law. Firstly, it was recognised that the current EU legal framework on copyright is sufficiently flexible and open to a new debate to bring into focus the challenges posed by questions of the copyright status of artificial intelligence as 'author'. The lack of harmonised rules on authorship of AI may lead to different solutions in the national law of different Member States with regard to works created with the help of AI, which could justify a future initiative to harmonise the legal framework in this respect.
Detailed consideration should be given to the collision between the notion of 'authorship as a human activity' and, on the other hand, artificial intelligence as a software, self-learning algorithm, which leads to a multitude of copyright effects. In my opinion, the aforementioned contradictions should be considered and analysed in the light of the general presumption of authorship in the hypothesis of Article 5 of the Directive 2004/48/EC on the enforcement of intellectual property rights. That provision enshrines the peremptory view that, in order for an author of a literary or artistic work to be regarded as such, in the absence of evidence to the contrary, and therefore to be entitled to bring infringement proceedings, it is sufficient for his name to appear on the work in the usual way. Importantly, the presumption of authorship in Directive 2004/48/EC applies with full force to works of artificial intelligence. Last but not least, the distinction between the notions of creation and generation of an object of copyright, sui generis, is the cause of the legal debate regarding the status of AI machine-generated databases, which may justify a revision or clarification of these notions in Directive 96/9/EC on the legal protection of databases.
4.Conclusion.
The debate on the copyright status of works created by artificial intelligence is still to come, and I am convinced that it will lead to interesting technological and legal developments, both in Europe and internationally.
The monumental advances in computer technology and the sheer amount of computing power available may make the 'human author'-'machine author' distinction meaningless. The fact that machine learning has yet to develop leads to the conclusion that humans will increasingly be replaced in the creative and scientific process by technologies based on artificial intelligence. Even today's reality points to sufficient computational arrays for self-learning computer systems, which may soon lead to a lack of distinction between content created by humans and content created by machines.
I think in this context that a balance should be sought in the understanding of "authorial contribution" as granting copyright to the person who made possible the functioning of the artificial intelligence that created a particular work, I find most logical, at least at this stage. This substantive copyright model already operates in the UK, and its enforcement leads to legal and technological efficiencies, as such an approach will ensure that companies will continue to invest in AI technology knowing that they will get a return on their investment. Perhaps the next big debate knocking at the door is whether AI workers will be granted the status and rights of persons analogous to the legal, social and psychological realities typical of human beings, but I'll leave that topic open for now.
Author: Mr.Atanas Kostov – attorney at law
[1] This is one definition of the term. Others include the understanding that "artificial intelligence is a field of study in computer science, arguing that artificial intelligence is concerned with the development of computers capable of human-like thought processes such as learning, reasoning, and self-correction" - so Joost N. Kok, Egbert J. W. Boers, Walter A. Kosters, and Peter van der Putten, "Leiden Institute of Advanced Computer Science", Leiden University, the Netherlands;
[2] MusicLM's musical works, include five-minute pieces created from just one or two words in the form of melodic techno, as well as 30-second themes that sound like entire songs and are formed musically into a specific genre, tempo, and specific, aurally recognizable instruments;
[3] Thus, Feist Publications v. Rural Telephone Service Company, Inc. 499 U.S. 340 (1991))";
[4] The U.S. Copyright Office's decision was made after a copyright review of the graphic novel Dawn of Dawn. According to the office, copyright protection was erroneously granted to the book's author, Chris Kashtanova, because the original copyright included all of the images in the book, which were generated using Midjourney AI software. Although the images cannot be copyrighted, USCO states that Kashtanova can copyright the parts of the book that she wrote and the way the AI-generated images are arranged on the page.
[5] See Australian case Acohs Pty Ltd v Ucorp Pty Ltd 2012. The case concerned a copyright dispute over HTML source code that was prepared by a computer program and included manually entered data;
[6] Thus ECJ Case No C-5/08 Infopaq International A/S v Danske Dagbaldes Forening;
[7] See CJEU Case C-145/10;
[8] This is CJEU Case No C-5/08 Infopaq International A/S v Danske Dagblades Forening;