Project Topics
Get Complete Project Material File(s) Now! »
Method and Material
The method that will be used in order to explore the legislative challenges of AI in the current IP framework is an EU legal dogmatic method. The analysis will be executed by examining relevant sources of law, in this thesis mostly EU law sources, but also some international law and news articles from around the world.4 For example European and international conventions such will be included. The EU law sources will consist of EU directives and regulations, along with reports from the European Commission, Council and Parliament. When examining the challenges on the current framework (the first question in chapter 1.2 above), primary and secondary EU-law sources and to the extent possible EU case-law will be consulted.
Famously in EU law the method is to interpret and analyse teleologically, that is, regarding what the purpose that is to be achieved is.5 Typically the preparatory work for EU law is also relevant for the EU method and interpretation.6 Therefore, regarding the two latter core questions in chapter 1.2, the analysis will be guided by the purpose behind IP protection. Furthermore, when analysis the solutions a selection of case law that shows different possible solutions to the problems that I am analysing will be presented. The intention in this part of the contribution is to explore the alternative solutions that can be identified and analyse how they would suit in the EU context by applying the general principle of promotion of innovation. When reading the part on how other legal frameworks have handled the issue under chapter 5.6, keep in mind that the information that I will use as inspiration is retrieved from secondary sources, i.e., other author’s description of a certain phenomenon and is not my own expert knowledge. The reason for my selection of case law from the United Kingdom (the UK), the United States of America (the US), Australia and China, is to examine reasonings from large states and economies, comparable with the EU. The reports from EU organs that will be used are intended to inform on possible drafting solutions that the legislator might make as part of my legal analysis. The purpose of this is to reach my conclusion by drawing from EU legislative custom, in addition to speculation from the legislator as well as from myself.
As AI is still a rather undefined area that can change rapidly it is currently mainly discussed (outside of policy documents) in scientific articles, and strong case-law has not yet been developed. Where scientific literature cannot be found, news articles will be the main source of information.
Limitations
IP rights derive from a vast number of different frameworks and protection regimes. This thesis aims to only examine the EU frameworks and will restrain from examining national regimes (with the exception of chapter 5 where inspiration from foreign national regimes will be made). Due to limited space and in order to achieve a consistent analysis, a fair selection of protection regimes will be made to illustrate the challenges on the IP system. For this reason, the regimes in focus will be copyright, design, trade mark and patent. The emphasis of the discussion will furthermore concentrate on the two regimes, copyright and patent, since these have gained most attention in the debate.
Disposition
The disposition of this thesis will be the following. First, an introduction to the topic of this work will be made in chapter 2 where some practical examples of AI as a creator will be presented. Chapter 3 will present the current IP framework in the EU, what IP rights aim to protect and how the protection is attained for each individual right. The next part, chapter 4, will examine the challenges on the IP framework caused by AI as a creator of subject-matter. Different solutions to the challenges on the IP framework will be analysed in chapter 5 and finally, an attempt to find the most suitable solution will be made in chapter 6. The two last mentioned chapters will contain the main analysis in this thesis. The result will be presented in the final chapter as well as the answers to the questions posed in chapter 1.2. In the very last chapter, 7.2, my proposal on the way forward for the EU legislator will be presented.
AI as a creator
In order to analyse the challenges on the current IP framework it is vital to address the state of the art of technology, and the potential of AI as a creator. Firstly, the object of this thesis – what a creator in the AI context is – will be examined. Therefore, definitions of AI will be given, followed by some practical examples of AI-generated art and other works.
Definition of AI
There have been various attempts to define AI.7 The World Intellectual Property Organisation (WIPO) defines AI in the context of IP as a “discipline of computer science that is aimed at developing machines and systems that can carry out tasks considered to require human intelligence, with limited or no human intervention”.8 The European Commission (the Commission) has in its proposal for a regulation on AI9 also attempted a uniform harmonised definition to provide legal certainty, although wide enough to include future technological developments.10 Article 3(1) of the proposal defines “artificial intelligence systems” as: “software that is developed with one or more of the techniques and approaches listed in Annex I and can, for a given set of human-defined objectives, generate outputs such as content, predictions, recommendations, or decisions influencing the environments they interact with”. Annex I of the Act informs us that the techniques included in the definition are: machine learning approaches,11 logic- and knowledge-based approaches,12 and statistical approaches.13 The aim of AI research and development is in general terms to examine human intelligence and to apply this when programming machines to think and act in ways requiring cognitive ability typically only found in human intelligence.14
In this contribution, the Commission’s definition from the proposal will be used and the terms “AI” or “AI-machines” will refer to especially the machine learning and logic- and knowledge-based technologies.
AI as a creator
AI-generated creations become a problem when the AI is creating autonomously from the programmer, owner or the user of the AI. Thus, it is vital to distinguish between AI creations that are autonomously generated by the AI, and creations where the AI is only an assisting element in a human generated creation.15 In the sector of automatic programming, researchers, as well as the Commission, mean that truly autonomous machines are still in the far away future (if ever).16 In the legal and policy discussion however the term AI-generated creations seems to aim more at the lack of creative involvement from a human, rather than the autonomy of the AI-machine, leading to undeserved recognition if the human is named as holder of the IP rights.17 The difference between AI autonomously generated creations and creations where AI has been used only as a tool is rarely clarified in the literature.18 During the research of this thesis, I have not found a clear and useful explanation. If I were to attempt to explain it with my understanding of the subject I would say that a machine as a helping tool would be for example when a machine it is programmed to facilitate a process for the user, for example the autocorrect and suggestion functions in your phone when you write a message, while AI that is programmed to write a message or paper without the involvement of the person at all, would be an autonomous act.
Below follows some practical examples of AI-machines that have “autonomously” created subject-matter to provide an idea of the types of creations this thesis aims to analyse.
Practical examples
Paintings
A perfect example of AI as a creator is “the Next Rembrandt” painting. This was a research project19 to “revive” the famous 17th century painter Rembrandt van Rijn by analysing his entire art collection in order to recreate a new masterpiece with the typical Rembrandt technique and style. In other words, they built a vast database consisting of paintings and fed this input into an algorithm, with the Next Rembrandt painting as the output.20
Another similar example is from 2018 where a painting drawn by an AI using a method called GAN (generative adversarial network) was created.21 The collective behind the experiment go by the name “Obvious”.22 The AI-generated portrait depicted the fictional French churchman Edmond Belamy. The paintings imperfection and interpretation of a human painted portrait of another human demonstrates, according to the researcher, that “algorithms are able to emulate creativity.”23 Can it really be called art though? The fact that the painting was sold for $ 432,500 at an auction indicates the creation’s worthiness of attention and the importance to take AI on the innovative market seriously.
Moreover, an AI research conducted by Ahmed Elgammal25 performed what he described as a visual Turing-test.26 Art made by machines and art made by humans was randomly presented to (human) evaluators, followed by questions such as if the evaluators thought the art was made by a person or a machine, and how inspiring they considered it to be. The results showed that the human evaluators were sometimes even more inspired by machine-made art, than art made by humans. The system that was used is supposed to be more of a creative network, not a generative one.27
The holder of authorship of these machine-generated paintings from an artistic point of view, is however not clear. The researchers mentioned above argue that if the creator of the image should be the author, this would be the machine. However, if the author should be seen as the “one that holds the vision and wants to share the message” it would be the creators of the machine, i.e., the researchers.28 Elgammal even goes as far as to suggest that the image is a product of the collaboration between two creators, one human and one machine.
Also, AI has been able to autonomously create music.29
Inventions
Regarding the area of patent AI is starting to invent things more autonomously. For instance, a food container based on fractal geometry was recently created by the DABUS30-machine. It has also invented devices, such as a type of flashlight, aimed to attract attention.31 The owner of the AI-machine, Stephen Thaler, applied for patent naming the machine as “inventor” to several patent offices around the world.32 This machine that has been programmed to autonomously invent has caused debate all around the world. Further details on this case and a new judgment from Australia will be discussed later under chapter 5.6.3.33 There are also older examples of AI inventions such as the Nasa antenna.34
Design furniture and trade mark
As mentioned above, the focus of the debate on AI as a creator for now revolves around copyright and patent. However, it is not hard to imagine that also design rights and trade mark for AI-creations could become just as relevant. To prove this, there are already examples of AI designing furniture35 and especially in the graphic design field, AI could come to be a huge threat to human graphic designers.36
Table of contents :
1 Introduction
1.1 Background – AI and IP law
1.2 Purpose
1.3 Method and Material
1.4 Limitations
1.5 Disposition
2 AI as a creator
2.1 Definition of AI
2.2 AI as a creator
2.3 Practical examples
2.3.1 Paintings
2.3.2 Inventions
2.3.3 Design furniture and trade mark
2.3.4 Summary
3 Intellectual Property Rights
3.1 Introduction
3.2 Copyright
3.3 Design rights
3.6 Recent EU proposals and reports
4 Challenges
4.1 Introductory remarks
4.2 Challenge 1 – Human requirement for protection
4.3 Challenge 2 – Harmonised regulation
4.4 Challenge 3 – Discouraging creation and monopolising
4.5 Challenge 4 – Derivative output
5 Possible solutions
5.1 Introductory remarks
5.2 Keep current framework unchanged
5.3 Legal personhood
5.4 Special computer-generated works regime
5.5 Sui generis authorship
5.6 Aspects and inspiration from case-law outside the EU
5.6.1 Copyright to creations that are made partly by a human and partly by a machine
5.6.2 Copyright for other non-humans, like animals?
5.6.3 Patent for machines – DABUS
5.7 Some ending comments
6 Analysis
6.1 Which is the most suitable solution for the EU framework on IP rights?
6.1.1 Unitary framework
6.1.2 Discussion between the presented solutions
6.2 Final remarks
7 Concluding statements
7.1 Result
7.2 The way forward
8 Bibliography
