INTRODUCTION

Artificial intelligence, according to the definition made by the European Commission, can be defined as “systems that display intelligent behavior by analyzing their environment – with some degree of autonomy – and taking actions to achieve specific goals”. So much so that it is argued by scientists that artificial intelligence is humanity’s last invention, and that all subsequent inventions will be carried out by artificial intelligence systems.

Determining the legal status of artificial intelligence is of great importance in terms of the nature of artificial intelligence products, the outcome of legal transactions established with artificial intelligence, or determining liability for the actions of artificial intelligence. Within the scope of our study, under the property status view, slave status view, and person status view; the natural person-like status view, artificial agent view, and electronic person view will be examined.

1. ARTIFICIAL INTELLIGENCE AND THE CONCEPT OF AUTONOMY

Today, humanity lives in a digital age alongside the rapid development of artificial intelligence and robotic technologies. Artificial intelligence technology, on the other hand, guides technology with the aim of producing a thinking machine similar to the human mind. We can define artificial intelligence as computer systems capable of performing tasks that require human intelligence.

The concept of autonomous vehicle refers to vehicles that can make decisions independently, autonomously, without external intervention, and execute these decisions. It is of great importance to determine the boundary of action and transaction capacities for robots with artificial intelligence to be granted personality. Even though traces of human input can still be found in fully independent autonomous systems, it is not entirely possible to say that the behaviors exhibited by autonomous systems are the result of consciousness. What is legally significant is that the behaviors of autonomous systems are undetermined and unpredictable.

Looking at the autonomy levels and development levels of artificial intelligence, artificial intelligence is divided into three. These appear as narrow (weak) artificial intelligence, general (strong) artificial intelligence, and super artificial intelligence. In addition, this division can also be made as narrow (weak) artificial intelligence, general (strong) artificial intelligence, theory of mind artificial intelligence, and super artificial intelligence.

Narrow (weak) artificial intelligence is a type of artificial intelligence that can specialize in a single subject and only answer questions in a specific area, and the artificial intelligence named Deep Blue belonging to IBM Corporation, which defeated world chess champion Kasparov, can be shown as an example. General (strong) artificial intelligence is an artificial intelligence that does not only react, but can make decisions, execute the decisions it makes, and manage processes at the same speed as a human. Self-driving vehicles or digital assistants can be shown as examples of this type of artificial intelligence. Lastly, super artificial intelligence is a type of artificial intelligence that can perceive human emotions and thoughts, establish social relationships with humans, possesses superhuman intelligence, can answer every question asked, and has performance superior to humans. There is no example of this type of artificial intelligence in real life yet, and for now it only appears in science fiction movies.

Within the scope of our study, while determining the legal status of artificial intelligence, narrow and general artificial intelligence types will be examined. Indeed, as there is no example of super artificial intelligence today, the proliferation of such artificial intelligence is not foreseen in the near future. In any case, there is no doubt that such a type of artificial intelligence could represent itself.

2. LEGAL STATUS OF ARTIFICIAL INTELLIGENCE

With advancing technology, human interaction with artificial intelligence has increased, bringing forth many legal questions. For example, questions such as what the fate of a contract concluded by artificial intelligence will be, who will bear liability for damages caused by artificial intelligence, and who will hold rights over things produced by artificial intelligence have rendered the determination of the legal nature of artificial intelligence necessary. Another question regarding artificial intelligence is who will be liable for its tortious acts. In order to answer these questions, the legal nature of artificial intelligence must first be identified.

Due to the absence of explicit regulations regarding artificial intelligence in our law as of yet, there are different views regarding the legal status of artificial intelligence.

2.1. The Property Status Argument

A The argument that artificial intelligence holds property status stems from the idea that artificial intelligence is a tool used for the benefit of humanity. According to this argument, artificial intelligence can be the subject of ownership. Based on this foundation, it relies on the idea that artificial intelligence is currently used as a tool and is not responsible for its actions, and accordingly, liability in disputes arising from artificial intelligence should be attributed to the owner.

In the developing world, in line with concepts such as the autonomy and unpredictability of artificial intelligence, it is also clear that artificial intelligence cannot be granted a simple property status. In order to discuss the property status of artificial intelligence, it is useful to first understand the concept of property.

In Turkish law, although property can be defined as “a corporeal, bounded thing over which control can be established”, in a broader sense, it can be defined as “impersonal corporeal entities over which control can be established and which have defined boundaries”. According to some authors in legal doctrine, another element of property is that it must possess economic value. However, according to the argument we adhere to, even if entities do not carry value in a material sense, those to which moral value can be attributed can also be defined as property; thus, it can be stated that property does not require an element such as carrying economic value.

The element of corporeality refers to the entity possessing a tangible physical presence, being a touchable three-dimensional object. It is clear that artificial intelligence does not carry the element of corporeality on its own. However, since artificial intelligence is essentially a computer program, similar discussions regarding the corporeality of computer programs can also be made for artificial intelligence.

In this context, according to one argument, computer programs acquire the element of corporeality indirectly and are considered property because they do not make sense without being loaded onto a data carrier, and computers do not make sense without the programs inside them. According to another argument, it is accepted that a computer program is an incorporeal entity upon which its programmer holds intellectual property rights. Indeed, the second argument has been adopted in Turkish law, and computer programs have been accepted under the status of works within the scope of Article 2 of the Law No. 5846 on Intellectual and Artistic Works.

Categorizing artificial intelligence as property is criticized most in terms of the element of impersonality. Indeed, it should not be possible to equate artificial intelligence programs, which are becoming increasingly intelligent, with a kitchen robot at home. When completely autonomous and intelligent artificial intelligences are concerned, it is highly likely that artificial intelligence may develop in a way that cannot be predicted even by its own programmer. As an example of this, the artificial intelligence named Tay, belonging to Microsoft software, learning to be racist and sexist through its chats with Twitter users, and even Microsoft being forced to shut down Tay because this situation reached a dangerous level, can be shown. Even this incident alone demonstrates how easy it is for artificial intelligence to get out of control independently of its creator, and this strengthens the claim that it does not carry the element of impersonality. Some authors argue that this situation that happened to Tay was not a case of getting out of control, but rather that Tay learned whatever was given to it, just as a young child would execute whatever you give them, and they state that what should actually be emphasized with this incident is that its actions are unpredictable considering the learning speed of artificial intelligence.

The idea of considering artificial intelligence as property could not be adopted by us either, as it fails to answer legal questions and cannot keep up with developing technology.

2.2. The Slave Status Argument

Although slaves in Roman law were considered property (res), in certain cases they exhibited the characteristics of a person (persona) and could engage in legal transactions. According to Locke, people should be able to hold property rights over the products they produce. For this reason, artificial intelligence produced by humans, even if it carries human characteristics, should have a status suitable for establishing ownership over it. This brings to mind the possibility of equating the status of artificial intelligence with that of slaves who were accepted under property status. According to this argument, rather than reviving slavery as advocated by the slavery view, the goal is to examine solutions to similar past issues through a historical approach.

In Roman law, a slave was in the status of property (res) under the ownership of their master, which could be bought, sold, or loaned. Slaves did not have legal capacity, but provided that certain conditions were met, they possessed capacity to act. A system was established where masters provided their slaves with a certain amount of capital, enabling slaves to enter into legal transactions. This capital given to slaves was called “peculium“. Although the authority to operate this money belonged to the slave, the capital legally remained with the master.

The concept of digital peculium can be defined as allocating a fund to artificial intelligence to operate as if it were its own. Advocates of this idea state that it opens a door to making artificial intelligence accountable without the need for any legal personality.

According to the argument we also adhere to, although the digital peculium proposal is a practical solution, attributing slave status to artificial intelligence should not be possible under modern world conditions. Roman law can serve as a guide to us. However, it is not correct to say that using the concept of peculium is only possible through a system of slavery. In this context, although the concept of peculium could be taken as a role model in future regulations, the argument that this should be done through a system of slavery cannot be adopted.

2.3 The Person Status Argument

Futurists argue that with technological developments, human knowledge will increasingly be transferred to networks in the future, and as a result, machines will also be able to understand, synthesize, and evaluate all of this information, thereby eliminating the distinction between human and machine.

Isaac Asimov has three laws of robotics. However, as robots develop their levels of autonomy, they may decide no longer to comply with these rules. Furthermore, it is clear that Asimov’s laws will not always be applied by artificial intelligence. Namely, even if artificial intelligences designed for purposes such as military espionage, dropping bombs, carrying military loads, etc., are beneficial for the society holding them, since it is inevitable that they will cause harm to the opposing side in a war environment, the thought that artificial intelligence will never violate Asimov’s laws would be far from reality.

Such being the case, whether robotic artificial intelligences—which possess the capacity to learn and develop, can make their own decisions, carry out human actions, have unpredictable actions, and hold the capacity to engage in legal transactions—can be considered persons will be examined under this heading.

2.3.1. The Human-Equivalent Legal Person Argument

A The argument advocating that artificial intelligence should have a status equivalent to humans argues that the defining feature separating humans from other beings is the ability to think and make decisions independently, and that with advancing technology, artificial intelligence will also be capable of doing so. Furthermore, this argument supports this idea on the grounds that artificial intelligence could be held liable when it makes a mistake or could be rewarded when it does something right.

However, it is certain that artificial intelligence has not yet reached a level of autonomy that would allow it to be considered in the status of a legal person. For example, in the event that the capacity of artificial intelligence is restricted due to reasons such as software errors, since the software error can be fixed, or if unfixable, the artificial intelligence can be deactivated, it should not be possible to evaluate artificial intelligence in the status of a fully autonomous legal person independent of external factors.

As we have noted, the argument advocating granting artificial intelligence a status similar to a legal person equivalent to humans bases this idea on artificial intelligence possessing the power of discernment and a conscious will. However, this reason alone will not suffice to grant artificial intelligence a status similar to a legal person.

2.3.2. The Legal Entity Status Argument

A The argument advocating that artificial intelligence should possess legal entity status argues that only a human can be a natural person, but this situation does not suffice to categorize artificial intelligence as property or a slave; that the relationship between artificial intelligence and the person managing it is similar to the relationship between a legal entity and its organ; and that, just as with legal entities, artificial intelligence can acquire personality by being registered in a registry.

This argument is criticized for asserting that there will be a relationship between artificial intelligence and its user similar to the relationship between a legal entity and its organ. Indeed, while the relationship between a legal entity and its organ is addressed through the concept of a “common purpose,” no such common purpose exists between artificial intelligence and the person who manages or uses it. Furthermore, this argument is criticized on the grounds that, in the event of liability on the part of artificial intelligence, determining the penal/monetary sanction as compensation limited strictly to the assets allocated to it would not be deterrent.

The stated criticisms are also adopted by us, and we are of the opinion that even if artificial intelligence is to possess personality, this can be neither natural personality nor legal entity status. Should granting legal entity status to artificial intelligence be in question, the question comes to mind whether a new company specifically allocated to artificial intelligence would be established. In such a scenario, while the company in the position of the producer of the artificial intelligence already exists, it is clear that establishing a company for artificial intelligence from scratch and appointing separate organs to the company would only further complicate matters.

2.3.3. The Artificial Agent (Representative) Argument

According to this argument, artificial intelligence should be evaluated within the scope of an agency relationship as a representative of a human. In other words, artificial intelligence is obligated to comply with the instructions of the person it represents, and if it fails to comply, it will be liable to the parties to the contract.

Although the artificial agent argument appears to be a practical way to resolve legal issues from a legal standpoint, since the artificial agent would still need to possess some form of person status in order to fulfill its liability toward the person it represents, this argument is not deemed appropriate by us.

2.3.4. THE ELECTRONIC PERSON STATUS ARGUMENT

The idea of granting electronic personality to artificial intelligence was proposed by the European Parliament Report with recommendations to the Commission on Civil Law Rules on Robotics. This idea is based on the necessity of introducing a specific legal status for robots. Thereby, it aimed to equip autonomous robots with specific rights and obligations, enabling the determination of their liability for the decisions they make and the damages they cause. The proposal envisaged the creation of a fund belonging to the autonomous robot, registering this fund with an individual registration number in an European Union register, covering damages resulting from the autonomous robot’s actions from this fund, and simultaneously carrying out financial transactions for the benefit of the robot through this fund.

This proposal was not adopted by the European Parliament. Indeed, as stated within the scope of the recommendation report, granting electronic personality would depend on artificial intelligence/robots being capable of making smart, autonomous decisions and interacting independently with third parties. Since this is not yet possible with today’s technology, the idea of granting an electronic personality to robots is not embraced by the European Parliament for now.

Although the electronic person argument is an acceptable approach in terms of registering artificial intelligence in a registry, thereby ensuring its transparency and enabling the determination of liability, it fails to answer questions regarding the fund to be granted to the artificial intelligence. Indeed, even who will pay for this fund cannot be determined.

2.4. The Work Status Argument

In Turkish law, the view that a computer program is an incorporeal entity upon which its programmer holds intellectual property rights has been adopted, and under Article 2 of the Law No. 5846 on Intellectual and Artistic Works (“Law 5846”), computer programs are accepted under the status of works. In legal doctrine, based on the fact that artificial intelligence is essentially a computer program, there is also an argument suggesting that it should be evaluated within the scope of intellectual property law.

The counter-argument, on the other hand, contends that under Law 5846, only natural persons can create works, and that artificial intelligence can be manufactured by both natural persons and legal entities; therefore, products created by artificial intelligence do not possess the quality of a work.

Besides this, another debate revolves around the question of whether artificial intelligence can be considered the author of the works it produces. In legal doctrine, the prevailing argument asserts that an entity to which personality has not been attributed under fundamental laws cannot be granted personality via Law 5846, and thus artificial intelligence cannot hold the title of author.

However, at this point, it is necessary to examine whether the products created by artificial intelligence can possess the quality of a work. Products generated by artificial intelligence (paintings, music, videos, scripts, games, etc.) lack originality and appear before us as imitative works. In other words, these products are mostly content that imitates or derives from prior data. Such being the case, it is clear that artificial intelligence cannot possess the status of an author.

In this context, there is no legal obstacle to attributing work status to artificial intelligence under today’s legal system. In my opinion, the conclusion has been reached that artificial intelligence possesses work status under Turkish law.

3. EVALUATIONS

Artificial intelligence can neither be categorized as property in the classical sense nor as a natural/legal person. Today, artificial intelligence has not yet acquired full autonomy. Furthermore, since artificial intelligence is essentially a computer program, it can be categorized as an intellectual work.

However, the main issue is not defining the legal nature of artificial intelligence, but rather determining how and to whom the liability arising from its actions will be attributed. What needs to be done is to establish the infrastructure that will ensure traceability and the transparent identification of liability in practice.

It has been concluded that there is no need to transform artificial intelligence into a subject within the legal order or to create a new personality status in its favor. Instead, it will suffice for every vehicle or robot possessing artificial intelligence to be registered in a central registry. Through this registry, all data that will be effective in determining liability—such as the ownership status, developer, programmer, user, and insurance information of the artificial intelligence—can be recorded regularly. Thus, in acts caused by artificial intelligence, it will become easier for injured third parties to identify whom to deal with, and legal uncertainty will be eliminated.

Our current legal system has regulated strict liability cases, such as the liability of the employer or the liability of the animal keeper; similarly, it is possible to evaluate artificial intelligence as a tool and resort to the objective liability of certain persons. In this context, a central registration system will serve as evidence in potential legal disputes regarding artificial intelligence and facilitate administrative oversight. Thus, a practical solution to the liability problem will be offered without the need for radical changes to the existing legal system. Thanks to the registry, to whom the artificial intelligence belongs, by whom it was developed, and by whom it is used will become clearly traceable, making it easier to determine the counterparty in the event of potential damage.

However, it should also be noted that the validity of this approach is limited to the conditions of the technology age we are currently in. Of course, with the evolution of technology, a regulation in this direction may be needed in the future; however, as of today, the idea has been adopted that there is no need for such a legal status and that what truly needs to be resolved is the determination of liability regarding artificial intelligence.

It has also been identified within the scope of our study that the main issue to be examined pertains to liability. Indeed, there are concrete questions that the legal world struggles to answer, such as who will be held liable in an accident involving a driverless autonomous vehicle, or to whom liability will be attributed if a smart home uses disproportionate force against an intruding burglar.

Following the rejection of the proposal to grant electronic personality to artificial intelligence by the European Parliament, focus was directed toward the determination of liability in incidents where artificial intelligence is used as a tool, and in this regard, the European Union Artificial Intelligence Act (AI Act – Regulation (EU) 2024/1689) was adopted.

With this regulation, obligations regarding risk management, data quality control, cybersecurity controls, transparency, and registering the AI system in the EU database have been imposed on the provider. For the user, it has imposed obligations such as using AI systems strictly in accordance with instructions, and monitoring and supervising the training of the AI. In this context, that the user will be directly liable for their use of the AI is another result of the regulation. Furthermore, within the scope of the regulation, the requirement for a fund to be provided by the provider has been regulated, just as in the electronic person argument. However, it should be noted that although the fund here will be used for the purpose of compensating damages caused by artificial intelligence, liability will still fundamentally belong to the provider.

As seen upon examining the regulation, it has been observed that the European Union has for now set aside discussions regarding the personality of artificial intelligence, and has attempted to regulate a strict liability regime where the provider and the user will be held liable for the liability arising from the use of artificial intelligence, just as in the liability of a vehicle operator.

Although the legal acceptance of artificial intelligence under the status of an intellectual creation was in question within the scope of our study, it has been determined that the main problem that needs to be resolved in artificial intelligence law pertains to liability law.

CONCLUSION

While the property status, slave status, and person status approaches examined in this study offer various possibilities for subjecting artificial intelligence to legal regulation, it has been observed that each of them possesses serious structural deficiencies.

It has been determined that computer programs possess work status under Turkish law and that artificial intelligence, being essentially a computer program, must also be considered under work status. It has been identified that artificial intelligence should possess a unique, distinct legal status in terms of accurately defining the liabilities arising from its activities and guiding these systems in line with the principles of human rights, safety, and social benefit.

Finally, although within the scope of our study it has been addressed that artificial intelligence should be evaluated under work status in Turkish law, it has been concluded that the main issue lying at the center of legal debates regarding artificial intelligence, as also stated within the scope of the European Union Artificial Intelligence Act, consists of problems related to liability law.

BIBLIOGRAPHY

Akbilek, Meliha. “Teknolojinin Pandora Kutusu: Robotların Hukuki Kişilikleri ve Hukuki Sorumlulukları”, Beykent Üniversitesi Hukuk Fakültesi Dergisi, C. 3, S. 6, Aralık 2017, s. 215-236.

Akkurt, Sinan Sami. “Yapay Zekanın Otonom Davranışlarından Kaynaklanan Hukuki Sorumluluk”, Uyuşmazlık Dergisi, Haziran 2019, S. 13, s. 39- 59.

Akkurt, Sinan Sami/ Abdurrahim Altun. “Yapay Zekaya Elektronik Kişilik Tanınması”, Ankara Üniversitesi Sosyal Bilimler Enstitüsü, Doktora Tezi, (915862), Ankara, 2023.

Bak, Başak. “Medeni Hukuk Açısından Yapay Zekânın Hukuki Statüsü ve Yapay Zekâ Kullanımından Doğan Hukuki Sorumluluk”, Türkiye Adalet Akademisi Dergisi, s. 211-232.

Demir, Remzi. Yapay Zeka İmalatçısının Ürün Sorumluluğu, Ankara, 2023.

Doğan, Erdem. Yapay Zekanın Hukuki Statüsü ve Sorumluluğu, 1.B., 2022.

Dural, Mustafa/ Tufan Öğüz. Türk Özel Hukuku Cilt 2 (Kişiler Hukuku), B. 19, İstanbul, 2018, s. 215.

Gayretli Aydın, Seda. Yapay Zekanın Hukuki Sorumluluğu, 2. B., 2024.

Kara Kılıçarslan, Seda. “Yapay Zekanın Hukuki Statüsü Ve Hukuki Kişiliği Üzerine Tartışmalar”, Yıldırım Beyazıt Hukuk Dergisi, Temmuz 2019, S. 2, s. 363-389.

Karaca, Uğur/ Esra Karataş. “Yapay Zeka Tarafından Meydana Getirilen Fikri Ürünlerin 5846 Sayılı Fikir Ve Sanat Eserleri Kanununa Göre Korunması”, Maltepe Üniversitesi Hukuk Fakültesi Dergisi, Haziran 2022, S. 1, s. 17-50.

Karadeniz, Özcan. “Roma Hukukunda “Pecilium” Müessesi”, Ankara Üniversitesi Hukuk Fakültesi Dergisi, C. 25, S. 3, 1968, s. 179-1944.

Kılıç, Muharrem. “Hukuksal Aklın Transhümanistik Temsilleri Ve Ontorobotik Varoluş Formları”, Adalet Dergisi, Mayıs 2021, S. 66, s. 15-54.

Oğuzman, M. Kemal /Özer Seliçi/Saibe Oktay Özdemir. Eşya Hukuku, 15. B. İstanbul, Filiz Kitabevi, 2012.

Özbay, Ümit Vefa. “Dijital Peculium Kavramı”, Ankara Üniversitesi Hukuk Fakültesi Dergisi, Ekim 2021, C. 70, S. 3, s. 867-98.

Serozan, Rona /Baki İlkay Engin/Yeşim M. Atamer. Medeni Hukuk Genel Bölüm/ Kişiler Hukuku, B. 9, 2022.

Sirmen, Lale. Eşya Hukuku, Legem Yayınları, 11.B., Ankara, 2023.

Tahiroğlu Bülent, Belgin Erdoğmuş. Roma Hukuku Dersleri, İstanbul, 2014.

Umur, Ziya. Roma Hukuku Ders Notları, İstanbul, 1999.

Yavuz, Baturalp. “Yapay Zekanın Hukuki Statüsü”, Bursa Barosu Dergisi, Eylül 2021, S. 121, s. 63-70.

Zorluel, Mustafa. “Yapay Zekâ ve Telif Hakkı”, Türkiye Barolar Birliği Dergisi, C. 32, S. 142, Mayıs 2019, s. 305-356.