* Turkish English version of this article has been translated from the Turkish original by artificial intelligence.
With the development of science and technology in the world, new concepts and problems are entering our lives. In this article, we will examine the copyright infringements of artificial intelligence. To understand the subject, we will first explain some concepts.
The first of these concepts is artificial intelligence, which is an information system that “performs various human brain behaviors such as perception, memorization, emotion, reasoning, proof, recognition, understanding, communication, design, thinking, learning, forgetting, creation, etc. artificially through a machine, system, or network.” (i)
Artificial intelligence is divided into four categories:
The first type of artificial intelligence is called a reactive artificial intelligence. Reactive artificial intelligences analyze the data in the subject they are programmed in and calculate potential outcomes. Examples of this type of artificial intelligence include the chess artificial intelligence DeepBlue or artificial intelligences developed for such table top games.
The second type of artificial intelligence is called limited memory capacity. These artificial intelligences analyze the commands given to them and act accordingly. An example of this type of artificial intelligence is Apple’s artificial intelligence Siri. What distinguishes this type of artificial intelligence from the first type is that the calculated data is more complex. To explain, the data received by the first type of artificial intelligence is taught and makes choices within a limited area, such as a chess board; the second type of artificial intelligence primarily understands what the received data is and acts accordingly.
The third type of artificial intelligence is called theory of mind. The typical characteristic of this type of artificial intelligence is its ability to understand human emotions and thoughts and communicate with humans. This type of artificial intelligence is not yet typical in our lives. However, the famous human-like robot Sophia is being developed with the aim of being suitable for this type. However, there are still some obstacles, such as not being able to fully understand emotions and not being able to develop a suitable approach accordingly.
The fourth and final type is referred to as self-aware. The difference between this AI and humans is that it is not organic. It knows that it was developed by humans, but a human can only understand that it is artificial intelligence by communicating with it. (ii)
The second concept we need to know is the concept of machine learning. The basic principle of machine learning is a technique that focuses on the continuous use of data and algorithms and gradually increases its accuracy. To understand machine learning, it would be appropriate to examine an experiment by OpenAI. In the OpenAI experiment, the robots that did not know the rules of hide-and-seek were allowed to randomly move by giving commands such as run away from red and catch blue. After 2.69 million trials, the reds learned to chase after 8.62 million trials, the blues learned to escape by using a vehicle after 14.5 million trials, and the reds learned to chase the blues by using a vehicle. This way, progress is made. (iii) As seen, AI improves itself by repeatedly doing the task given. The fundamental learning of artificial intelligences like DeepBlue is in this form.
Another concept is deep learning. Like machine learning, deep learning is based on scanning data, but its fundamental difference from machine learning is that the scanned data is constantly expanding and hosting a lot of data already. This data is called big data. The best example of this is FaceApp’s app that gives almost perfect results over time. FaceApp is a photo editing app and one of its services is to age the person in the photo. When the app first came out, this aging process consisted of only whitening hair and adding a few lines to the face and was very amateur compared to human work. Today, AI has understood how old people look because of the many times the app has been used and people sharing their age, etc. when using the app, and can now do aging realistically and professionally. In this example, big data is human photos. As can be seen, the difference between machine learning and deep learning is the quality and quantity of data used as a source.
Another concept is the concept of personality. In Turkish civil law and modern civil laws, the concept of personality is divided into two: real and legal personality. Real people are individuals with legal capacity, except for exceptions, they are equal in having both rights and capacity to act. Legal entities, on the other hand, are organizations and independent property groups established in accordance with the law, as stated in article 47 of the Turkish Civil Code, “Organized groups of persons having their own existence for a specific purpose, and independent property groups”. As people interact with artificial intelligence, they have started to experience some harm caused by AI and its use, and have discovered the difficulty of resolving these harms through the rules of classical tort law and the inadequacy of current legal understanding. This issue raises the problem of responsibility for AI. (iv) The general rule in liability law is that a person is responsible for harm caused to others by their wrongful act. However, in some cases, a person can be held responsible even if they are not at fault, known as strict liability. However, they can be released from liability if they prove that they have taken the necessary precautions and exercised the required care.
There is no legal regulation on the responsibility of AI in most countries. Therefore, various opinions have been expressed:
The first opinion is the generally accepted one in Turkish law, that of machine and computer systems being classified as objects. However, this view is problematic because of the ability of AI to make decisions on its own.
The second opinion is to classify AI as a slave status. However, how appropriate would it be to attribute a status that was accepted centuries ago that now is not seen as a humain practice to the AI’s that we are yet failed to understand fully in this advancing times of technology?
The third opinion is that AI will have a legal entity. The legal entity view is consistent with the cognitive features and potential generated by the AI’s intelligence and cognitive capacity. Furthermore, the relationship between an AI and the person who created/manages it resembles the relationship between an association and its management board members. (v) However, this view is criticized for the lack of clear regulation of legal entities in the law and the inability of any AI to fit into this status.
Another opinion is to see AI as an assistant and to hold the person who employs the AI responsible. However, this view is also problematic because the employer may not always be in a position to control the actions of the AI and in order for artificial intelligence to be a party, it must be put into a certain legal status already.
The last view is the electronic personality model proposed by EUROROBITICS. The electronic personality model, based on the idea that artificial intelligent entities should have a status different from that of objects, is based on the idea that a similar status can be granted to artificial intelligent entities, taking into account the legal status of legal entities. (vi)
After our statements about the legal status and responsibility of artificial intelligence, we will discuss the issue of whether artificial intelligence violates the rights of artists.
The starting point of the problem is Midjourney, an artificial intelligence that produces artistic images from entered text and images. Midjourney uses deep learning to analyze images on the internet and uses them for its programming purposes. Because the images that it analyzes are made by humans and have the signatures of the works owners, it uses signature-like lines in the images that it produces. This has made people worried about the violation of their copyrights.
When examining these allegations, legal grounds will be based on the following provisions of the Code on Intellectual and Artistic Works:
Article 1/B- Work is for all kinds of intellectual and artistic products. The owner of a work is for the one who made these of intellectual and artistic products.
Article 7- A work that is presented to the public with the consent of the right holder is deemed to have become public. If the copies of a work obtained by duplication from the original are offered to the public by being put up for sale or distributed or otherwise placed in a commercial location with the consent of the right owner, that work is deemed to have been published.
Article 8- The owner of a work is the person who creates it. The owner of a processing and compilation is the person who processes it, provided that the rights of the original work owner are protected.
Article 16- Without the permission of the work owner, no alterations, additions, or other changes may be made to the work or in the name of the work owner. The person who processes, reproduces, represents, or otherwise disseminates a work with the permission of the law or the work owner, may also make alterations required by the processing, reproduction, representation, or dissemination technique without the special permission of the work owner.
Article 21- The right to benefit from a work by processing it exclusively belongs to the work owner.
In addition to these, the important thing to know is that there is no need for registration for copyright to arise, the rights on intellectual and artistic works arise with the production of the work.
As can be seen, examining and analyzing works does not inherently violate copyright. In most cases, the owners of the works publicly publish a part of their works through intermediaries in order to obtain economic benefits and obtain business. Processing this data by violating the rights of the work owner may create a copyright infringement. However, Midjourney primarily does not process works, but rather presents something new inspired by them, so original works are only possible in exceptional cases. That is, the lack of legal barriers for artificial intelligence to process text input based on data drawn from open sources based on its mode of operation, except for exceptions. These exceptions also vary depending on the nature of the agreement or statements made between the work owner and publisher or during the publication of the work by the work owner.
So how can unpublished works be used by AI? There are two methods. The first is for the AI producer to program AI to analyze the unpublished work, which will inherently violate copyright because the deliberate violation of the rights of the work is involved. The second is for the AI user who somehow gains access to the unpublished work to process the work using AI. In such cases, the existence of the infringement is determined based on the purpose of the work, how it was used, and whether it caused economic harm to the original work owner. The only exception is parody of the original work, which is required by fair use rights. (viii)
There is currently a lawsuit opened due to user infringement. (ix) Upon the filing of the lawsuit, the producers of Midjourney, the defendant, announced that in the event of a lawsuit being filed against them for using works through their sites, they would reclaim the compensation they would pay from the user who used the work. As can be seen, the lawsuit was opened against the producer due to the user’s fault under strict liability. The lawsuit is still ongoing.
In conclusion, artificial intelligence is still a new concept and has not yet found its full equivalent in contemporary legal systems. Loss of rights caused by factors such as this and similar reasons can lead to lawsuits that are directed at the manufacturer due to the fact that artificial intelligence has not yet reached levels 3 and 4. Such lawsuits are initiated by adopting the principle of strict liability of the manufacturer through interpretation. The results of lawsuits initiated from the perspective of works of art vary according to the copyright, processing rights, and publicizing processes of the work owners.
You can access the presentation material used in in-office training on the same subject as this article here.
Uçar Law & Consultancy Office – Artificial Intelligence and Copyright – Presentation Material
All the images used in the presentation material are unique visual products created with artificial intelligence.
You can contact us here for detailed information on the subject.
You can get detailed information about the work of our office in the fields of Information Technology Law and Contracts Law on our website.
Ali AYDOĞDU (Student Intern)
Uçar Law & Consultancy Office
REFERENCES
- Deyi Li, Yi Du. Artificial Intelligence with Uncertainty. Boca Raton. CRC Press. 2016.
- Arend Hintze, Understanding the Four Types of AI, From Reactive Robots to Self-Aware Beings. Observer. 2016.
- Bowen Baker, Ingmar Kanitscheider, Todor Markov, Yi Wu, Glenn Powell, Bob McGrew, Igor Mordatch, Emergent Tool Use from Multi-Agent Interaction, OpenAI, 2019
- Başak Bak. Medeni Hukuk Açısından Yapay Zekânın Hukuki Statüsü ve Yapay Zekâ Kullanımından Doğan Hukuki Sorumluluk, TAAD, 2018
- Emre Bayramlıoğlu, Akıllı Yazılımlar ve Hukuki Statüsü: Yapay Zekâ ve Kişilik Üzerine Bir Deneme, Alacakaptan’a Armağan, İstanbul, 2008.
- Gizem Özkan Şahin, Çağatay Şahin, Yapay Zekalı Varlıklara Elektronik Kişilik Modeli Tanınmasına İlişkin Eurobotıcs Raporu Ve Fikri Mülkiyet Sorunu Bağlamında Meseleye Yaklaşım, İnönü Üniversitesi Hukuk Fakültesi Dergisi, Malatya, 2022.
- C. Kültür Ve Turizm Bakanlığı, Telif Hakları Genel Müdürlüğü İnternet Sitesi: Telif Hakları.
- Linda Joy Kattwinkel, Is Fan Art Legal Fair Use? What About Mash-ups? – Copyright Myths and Best Practices, OWE, 2021
- James Vincet, AI art tools Stable Diffusion and Midjourney targeted with copyright lawsuit, The Verge, 2023
