Ankara --:-- · —
bilgi@utuhukuk.com
0536 593 16 00
Yapay zekâ ve hukuk relationship is debated on many issues, from data protection to damage arising from the autonomous behaviour of artificial intelligence. In this article we discuss the definition of artificial intelligence, its legal development, its legal status and the civil and criminal liability arising from artificial intelligence.
Currency note
This article was prepared in accordance with the legislation and practice in force on its date of publication. New rules on artificial intelligence and product liability have been adopted in the European Union, and there have been significant changes since then with regard to the EU initiatives and proposals mentioned in the article. You are advised to consult a lawyer for the current position.
Many definitions of artificial intelligence have been put forward. These definitions contain broadly the same elements. To cite a few of them: according to one definition, “artificial intelligence: machines that have the ability to solve problems that humans solve with the intelligence they naturally possess”; according to another definition, “it is perceiving something complex and making appropriate decisions”. Although interest in artificial intelligence did not reach very high levels in the 21st century, it gained rapid momentum in its development after 2010.
With the development of technology, the concept of “data” has also started to gain importance. In parallel with technological developments, the processing, transfer and storage of a large proportion of the data held by individuals have come to the fore. These effects of technology have begun to appear in the form of applications that have become an ordinary part of our daily lives and that make our lives easier as they develop rapidly day by day, and they have gradually come to form an inseparable and important part of our lives. Unlike humans, artificial intelligence has the capacity to perform far more advanced computation and to process very large volumes of data. While offering many benefits, this technology, which makes the processing of data possible, has also rendered individuals vulnerable by gaining control over their data. Accordingly, as work on artificial intelligence has increased, its interaction with the field of law has increased in parallel. This is because the protection of individuals' data is a responsibility that the law must assume.
The Sophia example
Perhaps the most important event that brought artificial intelligence onto the public agenda was the news in October 2017 that Saudi Arabia had granted citizenship to Sophia, a social humanoid robot. This news was significant in two important respects: first its sensational aspect, and second its legal aspect. Whereas the concept of citizenship had until then been used only in respect of human beings, Saudi Arabia’s move brought onto the agenda the question: is it possible for the concept of citizenship to be used for non-human entities as well?
Another effect of artificial intelligence concerns private life and rights. With the development of artificial intelligence, debates have arisen on matters such as data protection and the liability or rights arising from artificial intelligence models. “This increases the importance both of understanding what technology brings to the field of law and of the need for the legal boundaries and obligations to be determined and known by the researchers developing the technology. Indeed, for artificial intelligence systems, law has a two-way relationship: both as a field in which it can produce solutions and in ensuring that the process for solutions produced in other fields operates in a controlled and reliable manner. It has an active relationship owing to the need for legal support throughout the process of generating, collecting and storing data and of the decision system produced by artificial intelligence”[2]. Ultimately, technology and law can no longer be considered separately.
The RoboLaw Project, supported by the European Parliament on 1 March 2012, was completed in May 2014, and the “Guidelines on Regulating Robotics” were published. The aim of the project was to examine the legal and ethical implications of robotics, bionics, neural brain–computer interfaces and nanotechnologies; to investigate whether the existing legal framework is adequate; if this research revealed shortcomings, to identify the changes needed; and to investigate their effects on the social values and processes that matter to people[3]. Following such projects and studies, certain legal instruments have been adopted in the field of artificial intelligence. Examples include Civil Law Rules on Robotics (EPLO 2015/2103), Legal and Ethical Reflections Concerning Robotics (EPRS, STOA Briefing, June 2016) and A European Perspective on Robot Law (July 2016). Perhaps one of the most important of these is that, on 10 April 2018, the Member States of the European Union signed the Declaration of Cooperation on Artificial Intelligence, formally committing to take measures that would place the European Union at the forefront of global competition in this field and to carry out scientific, legal and ethical work[4].
Studies and instruments mentioned in the article:
One of the most important features of artificial intelligence is its ability to take decisions and act autonomously. As a result of artificial intelligence taking decisions and acting autonomously, life, bodily integrity, health, privacy of private life, personal data, honour and dignity it may infringe personal values falling within personality rights, such as the above, and may also cause pecuniary and non-pecuniary damage as a result of certain economic and ethical errors[5]. Recently, the question of artificial intelligence used in the military field ending, through its autonomous decision, the right to life or harming health and/or bodily integrity has been much debated. Likewise, these risks may also arise from artificial intelligence used for medical purposes or from the misuse of driverless vehicles. As an example of artificial intelligence used for security and surveillance purposes, a dronemay record data which, if disclosed in some way, could result in a breach of the privacy of private life. Where artificial intelligence used in online shopping or online banking decrypts personal data, an infringement of personality rights and the occurrence of pecuniary/non-pecuniary damage are of course also possible. In order to address these problems arising from the autonomous behaviour of artificial intelligence, it is important first to establish the legal status of artificial intelligence.
For liability to arise from damage, the capacity to have rights and obligations, that is, legal personality, is first required. Accordingly, for artificial intelligence to be held liable for damage arising from its autonomous decisions, it must have personality under the Turkish Civil Code; no such status has been granted to artificial intelligence in Turkish positive law, and it is therefore not possible to speak of liability. However, there remains damage and an artificial intelligence to which liability cannot be attributed; in other words, the problem has still not been solved. In legal doctrine, regarding the legal position/status of artificial intelligence, “thing”, “slave”, “legal person” and “electronic person” and similar views have been put forward.
Views in legal doctrine and the assessment in the article
| View | Assessment in the article |
|---|---|
| Slave | Not possible, as the status of slavery is not recognised in the modern legal system |
| Thing | Considered inadequate in terms of liability, since artificial intelligence is not an ordinary thing |
| Legal person | Not considered technically possible, as there is no element of a common purpose; it gives rise to a problem of representation |
| Electronic person | Recommended by the European Parliament’s Committee on Legal Affairs (comparative law) |
Since the status of slavery is not recognised in the modern legal system, it is not possible to place artificial intelligence in the status of a “slave” . Under the view that it is a “thing”, artificial intelligence is regarded not as a subject of rights but merely as an object over which the property rights of subjects endowed with personality by law are exercised; however, owing to its cognitive abilities, artificial intelligence is not an ordinary thing and has the capacity to cause certain damage that perhaps not even a human being could cause, and therefore accepting the status of item is considered insufficient in terms of liability.
Those who advocate the legal personality view have stated that there is a relationship between artificial intelligence and the person who creates and/or manages it that resembles the relationship between an association and its board of directors. The definition of an association in Article 56 TCC refers to the element of “common purpose” , and it does not appear technically possible to speak of a common purpose shared by artificial intelligence and the person who creates/manages it. Furthermore, providing for personal liability of artificial intelligence for its acts by granting it legal personality and separate assets would also give rise to the problem of representation in actions brought against it. Therefore, rather than providing artificial intelligence with legal personality and separate assets, it would be more logical, at least in terms of liability, to hold those who create/manage it liable under the provisions on general strict liability for hazardous activities or under sui generis strict liability rules.
In comparative law, the electronic person view has been put forward. “The Committee on Legal Affairs of the European Parliament has stated that the unpredictability in the behaviour of the new generation of robots (artificial intelligence), which learn from their own variable experience and interact with their environment in a unique and unforeseeable manner, can be analysed and resolved only to a certain degree within the framework of the provisions of Directive 85/374/EEC, the existing legislation (APR.art.Z/ AI26), and has therefore set out the need to adopt a framework regulation laying down ethical rules on the matter (APR.art.1127). In this context, the Committee recommended the creation of a specific legal status of ‘electronic personality’ for robots (artificial intelligence) (in terms of liability arising from their autonomous decisions) (APR. art.59/ f28)”[6].
As mentioned above, since there is no legislative provision under which we could attribute to artificial intelligence the status of a legal person or any other form of personality, it will of course also not be possible to characterise it as a subject of liability. However, the question of which rules of positive law will govern how the persons who are to assume and/or share liability arising from the autonomous behaviour of artificial intelligence assume that liability gives rise to differing views.
Where artificial intelligence mediates the conclusion or performance of a contract and the injured party is the counterparty to that contract, the resulting “contractual liability” It lies, under Art. 112 et seq. of the TBK, with the person on whose behalf/account the artificial intelligence acts or the user. It is also possible that the strict liability of the person using artificial intelligence for the acts of an auxiliary in performance may simultaneously arise by analogy under Art. 116 of the TBK.
As a result of the autonomous behaviour of artificial intelligence, “non-contractual liability” may arise in various ways. First of all, it may occur to us whether the person using artificial intelligence can be held liable under Article 66 TCO, “strict liability of the employer” . However, since there is no employment relationship between artificial intelligence and the person using it, as required by Article 66 TCO, it is not possible to apply this article in terms of liability.
Another view is that the user of artificial intelligence is liable under Article 69 TCO in the manner of a “building owner” . However, for this provision to apply, according to the settled practice of the Court of Cassation, the structure causing the damage must be a “construction attached to the ground” such as a building, road, dam, embankment, electricity pole, grandstand, stadium and the like. As it is required that the structure be attached to land, and artificial intelligence does not have this element, it will of course not be possible to apply this article either.
TCO provisions discussed in non-contractual liability
| Provision | Assessment in the article |
|---|---|
| Art. 66 of the TBK — liability of the employer | Not applicable, as there is no employment relationship |
| Art. 69 of the TBK — liability of the owner of a building or structure | Not applicable, as artificial intelligence is not a construction attached to the ground |
| Art. 67 of the TBK — liability of the keeper of an animal | One of the reasonable approaches for application by analogy |
| Art. 71 of the TBK — liability for hazardous activities | The second reasonable approach; the tort provisions of Art. 49 et seq. of the TBK are reserved |
It can be said that there are two reasonable approaches to the matter. The first is that the user of artificial intelligence be held liable by analogy under Article 67 TCO, under the provisions on “liability of the keeper of an animal” . Article 67/1 TCO states that the person who permanently or temporarily undertakes the care and management of an animal is (strictly) liable for the damage caused by that animal. Legal persons, like natural persons, may also be held liable in their capacity as keeper of an animal. The keeper of an animal is to be understood as the person who permanently or temporarily undertakes the care and management of an animal.
The second approach is that the user of artificial intelligence be held liable under Article 71 TCO, within the framework of the provisions on “hazard liability” . Article 71/1 TCO sets out the principles of liability by providing that “Where damage arises from the activity of an enterprise posing a significant degree of danger, the owner of the enterprise and, if any, its operator shall be jointly and severally liable for such damage.” In determining whether an enterprise qualifies as an enterprise posing a significant degree of danger under this article, its “nature” or the nature of the “elements used in the activity (materials, tools or forces)” must be taken into account. Furthermore, where it is concluded that the enterprise is “apt to cause frequent or serious damage” even if all the care expected of a person has been exercised, the enterprise may be regarded as an enterprise posing a significant degree of danger. Article 71/2, first sentence, TCO “Where, having regard to the nature of a business or the materials, instruments or forces used in its activity, it is concluded that the business is liable to cause frequent or serious damage even if all the care expected of a person specialised in such work is exercised, it shall be deemed to be a business posing a significant hazard.” defines this as such. Consequently, whatever the activity of the enterprise, if it is concluded that, owing to the artificial intelligence (that is, the tool) used in carrying out that activity, it is apt to cause frequent or serious damage even where all the care expected of an expert in such work has been exercised, that enterprise may be held liable under Article 71 TCO for the damage suffered by third parties as a result of the autonomous behaviour of the artificial intelligence; in any event, however, the tort provisions set out in Articles 49 et seq. TCO are reserved.
“Where damage arises from the activity of an enterprise posing a significant degree of danger, the owner of the enterprise and, if any, its operator shall be jointly and severally liable for such damage.”
Turkish Code of Obligations, Art. 71/1
In the early days of artificial intelligence, it is observed that people used it as an instrument in committing offences. Later, however, it was claimed that, by means of certain software, they could think by themselves and engage in autonomous behaviour , and the question arose as to how the problems resulting from such autonomous behaviour of artificial intelligence should be resolved.
In 1981, at a company manufacturing motorcycles, a 37-year-old Japanese worker was perceived by a robot artificial intelligence working at the same company as a threat to the task it was aiming to perform; deciding that the most effective way of ending this threat was to eliminate the worker, the robot pushed the worker into an operating machine with its powerful hydraulic arm. As in this example, as artificial intelligence develops, it has begun to bring such problems to the fore. With regard to the attribution of criminal liability arising from such incidents, it is important whether artificial intelligence itself may be directly liable, apart from the person who produces or programs it. Therefore, the question who is the perpetrator? must be answered if justice is to be done in respect of such incidents. Under Turkish criminal law, only a natural person can be a perpetrator, whereas various sanctions are provided for legal persons. In determining who bears criminal liability, the fundamentals of criminal liability must be examined.
Maddi unsur
Occurrence of conduct matching the definition of the offence; examination of the conduct element of artificial intelligence
Manevi unsur
Free will; determining the extent to which the elements of knowledge and intent are present
For an act to qualify as an offence, it must contain both material and mental elements. For an offence to exist, conduct corresponding to the statutory definition must occur. The examination of the act element of artificial intelligence is important at this point. In the case of robots, this conduct element is generally visible. The conduct element may appear in different forms in relation to artificial intelligence.
Where an artificial intelligence managed by a human is concerned, the conduct element will have been carried out by the person managing the artificial intelligence, and, under Article 37 of the Turkish Criminal Code, persons who use another as an instrument in committing an offence will be held liable not as indirect perpetrators but as direct or sole perpetrators. In terms of the emergence of criminal liability, the acts of artificial intelligence will cause problems in establishing the causal link and in objectively attributing the result. There is no problem where artificial intelligence is used merely as an instrument. However, the resolution of situations in which artificial intelligence is involved through autonomous behaviour, or which it brings about through omission, will be complex.
The concept of free will possessed by human beings is important for liability in criminal law. As a rule, offences are committed intentionally; however, for certain offences the legislator has provided that they may also be committed negligently. The person must be aware of the act he or she has committed and of the consequences that act will produce. On the other hand, intent requires both knowledge and will. In this sense, it is important to determine whether artificial intelligence has free will, that is, to what extent it possesses the elements of knowledge and will. In cases where there is no free will, artificial intelligence will be used merely as an instrument, and it will therefore again be possible to resolve this problem in the light of current legal rules. However, if there is an artificial intelligence with free will and capacity for fault, it will also become necessary to determine the legal status of such entities.
Rümeysa ÖZYÜREK
[1] Eriş, M. Hakan, “Yapay Zekâ: İnsanların Doğal Olarak Sahip Oldukları Zekâ İle Çözdükleri Problemleri Çözme Becerisine Sahip Makinelerdir”, Yapay Zekâ Çağında Hukuk, Workshop Report of the Istanbul, Ankara and Izmir Bar Associations, 2019.
[2] Kızrak, M. Ayyüce, “Teknoloji, Bilimsel Araştırmalar Kadar Hukuki Düzenlemeye de İhtiyaç Duymaktadır”, Yapay Zekâ Çağında Hukuk, Workshop Report of the Istanbul, Ankara and Izmir Bar Associations, 2019.
[3] Ünsal, Burçak, “Yapay Zeka, Robotlar, Hukuki Düzenlemeler”, İstanbul Barosu Dergisi, July-August 2019, Vol: 93, Issue: 4, 69-70.
[4] Ünsal, p. 70.
[5] Akkurt, Sinan Sami, “Yapay Zekanın Otonom Davranışlardan Kaynaklanan Hukuki Sorumluluk”, Uyuşmazlık Mahkemesi Dergisi, June 2019, Year: 7, Issue: 13, pp. 43-44.
[6] Akkurt, pp. 46-47.
[7] Akkurt, Sinan Sami, “Yapay Zekanın Otonom Davranışlardan Kaynaklanan Hukuki Sorumluluk”, Uyuşmazlık Mahkemesi Dergisi, Year: 7, Issue: 13, June 2019, pp. 39-59.
[8] ÖZBEK, C./ ÖZBEK, V. Ö., “Yapay Zekânın Dâhil Olduğu Suçlar Bakımından Ceza Hukuku Sorumluluğunun Belirlenmesi”, Ceza Hukuku Dergisi, Vol: 14, Issue: 41, December 2019, pp. 603-622.
According to the article, artificial intelligence can process very large amounts of data, which gives rise to debates on issues such as personal data protection and the liability and rights arising from artificial intelligence models. The article therefore states that technology and law have become inseparable.
No. According to the article, legal personality is required in order to be held liable for damage; Turkish positive law does not grant artificial intelligence any such status, and therefore the liability of artificial intelligence itself cannot be spoken of.
The views that it should be treated as a thing, a slave, a legal person or an electronic person have been put forward in legal doctrine. The article considers that slave status is not accepted in contemporary law, that the status of a thing is regarded as inadequate in terms of liability, and that the legal personality view would create problems of representation.
According to the article, if the injured party is the other party to the contract, contractual liability under TBK Art. 112 et seq. lies with the person in whose name or on whose account the artificial intelligence acts, or with the user; the user's liability for the acts of an auxiliary person under TBK Art. 116, applied by analogy, may also arise.
Makalede istihdam ilişkisi bulunmadığı için TBK m. 66’nın, yeryüzüne bağlı inşa eseri niteliği taşımadığı için TBK m. 69’un uygulanamayacağı; buna karşılık TBK m. 67’nin kıyasen uygulanması veya TBK m. 71 tehlike sorumluluğunun akla yatkın yaklaşımlar olduğu görüşü savunulmaktadır. Haksız fiil hükümleri saklıdır.
According to the article, in Turkish criminal law the perpetrator can only be a natural person. If artificial intelligence is used merely as a tool, the person directing it is liable as the perpetrator under TCK Art. 37; the resolution of situations arising from the autonomous behaviour of artificial intelligence, however, becomes complex.
Consultation request
As UTU Hukuk Danışmanlık, we provide legal and advisory services in many areas of law.
Fill in the form and we will call you. You can also contact us directly: