Ankara --:-- · —
bilgi@utuhukuk.com
0536 593 16 00
With the entry of artificial intelligence into our lives, the legal status of artificial intelligence and the question of who will be liable for the damage it causes have come onto the agenda. As is well known, law must closely follow social developments, changing forms of governance and technology, because law has a dynamic structure and must keep renewing itself. In this article we discuss the definition and levels of artificial intelligence, the legal problems that may arise in the fields in which it is applied, the theories concerning its legal status, and contractual and non-contractual liability.
Currency note
This article was prepared on the basis of the legislation and practice in force on the date of its publication. There may have been significant developments since then in the legal regulation of artificial intelligence, particularly at European Union level. You are advised to consult a lawyer regarding the current position.
In order to understand what artificial intelligence is, it is useful first to consider the plain meanings of the words.
According to the Turkish Language Association (TDK), artificial: “Made or derived by human hand in imitation of its examples in nature; man-made, artificial, synthetic; the opposite of natural.” means. Intelligence then: “to think about events independently, to adapt successfully to new situations, and to bring actions and attitudes together around a particular idea or purpose.” This is what the word means. Intelligence is a phenomenon whose existence has been proven in many living species. Artificial intelligence, on the other hand, is a form of reasoning performed by machines and information systems. Artificial intelligence has a history of some 70 to 80 years.
Artificial intelligence is divided into three levels.
Artificial intelligence is present in many fields such as banking, shopping, commerce, communications, industry and information systems, yet for many people today the first thing that comes to mind is robots. From the perspective of artificial intelligence, a robot is a machine that operates by means of a computer program which enables it to coordinate the timing, intensity and number of repetitions of the sequence of movements it must perform, and even the activities of other robots, and which also contains safety definitions such as when or under what conditions it must stop. Following this definition of a robot, a question arises: which areas of the legal system will these machines equipped with artificial intelligence touch upon?
No field that could truly be called artificial intelligence law has yet been fully formed anywhere in the world. There is no doubt that this formation will be completed in the future. However, by turning the fact that our country lags behind other countries in technology into an advantage in this respect, it would be in the interest of our society for an artificial intelligence law to be created under the leadership of the legislator, with the assistance of experts in the field, by lawyers who have worked on this subject.
We mentioned that, with its entry into human life, artificial intelligence has come to be used in many fields. As its use has increased, this has also brought legal issues with it. These autonomous machines equipped with artificial intelligence may give rise to harmful situations as a result of their interaction with humans or with one another. To give some examples: if it is an autonomous vehicle, it may be involved in an accident and cause pecuniary and non-pecuniary damage, giving rise to both criminal and civil liability. It may be a computer program concluding contracts on behalf of a company and may make an error in the essential elements of the contract. If it is data-gathering software (crawling) and accesses information it should not access, liability may arise in relation to the protection of personal data. In view of all these situations, in order to understand the legal liability of tools using artificial intelligence, we must first determine whether artificial intelligence is a subject or an object of law.
The problems given as examples in the article are as follows:
In law, liability is specific to legal subjects. The subjects of law are persons in the legal sense, and under Turkish law persons are divided into natural persons and legal persons. This distinction is set out in the Turkish Civil Code (TCC). It is clear and unequivocal that natural persons are human beings. Natural persons, on the basis of their capacity to act, can hold certain rights and can assume obligations of their own free will. In this respect, the fact that artificial intelligence is not a natural person stems from its lack of capacity to hold rights and to assume obligations.
Turning to the theories on the legal status and personality of artificial intelligence, first:
Object (thing) theory
Artificial intelligence is not a subject of law but an object over which persons’ property rights may be exercised; however, its autonomous behaviour shows that it should be distanced from the classic definition of a thing.
Legal personality view
It is argued that artificial intelligence is not a thing and that personality should be attributed to it; since it cannot be regarded as a natural person, what remains is for it to fall within the concept of legal personality.
According to the theory that it is to be regarded as property, it is not a subject of law but an object over which persons’ property rights may be exercised. However, it is debatable how appropriate it is to bring it within the scope of property in the classical sense; admittedly, it is not a living being and it does not possess intelligence by natural means. On the other hand, by virtue of programming, it has features such as problem-solving, the capacity for reasoning and the display of autonomous behaviour. In this respect, the actions it performs and the behaviour it displays also indicate that it should be distinguished from the classical definition of property.
According to the views that it constitutes a legal person, artificial intelligence is argued not to be property and should be attributed a personality. This cannot be achieved within the concept of a natural person for ontological and epistemological reasons. What remains, therefore, is for it to fall within the concept of legal personality. Legal persons are entities consisting of people who come together around a particular purpose or of assets dedicated to a particular purpose, and which have a personality independent of the persons who bring them together. Legal persons also have legal capacity within the legal order.
When we examine whether artificial intelligence can be a legal person, there must first be a purpose, but this is not really possible for artificial intelligence, because a unity of purpose between artificial intelligence and the persons who create or manage it is technically impossible. Furthermore, if artificial intelligence were granted personality, it would itself have to be liable for its autonomous behaviour. In order to compensate for pecuniary damage it has caused, it would need to have assets of its own. This would bring other legal problems with it and would currently have no legal basis under Turkish law.
Comparative law · Electronic person
As a concrete piece of work in this field from the perspective of comparative law, the European Parliament’s Committee on Legal Affairs included recommendations in the report it published in 2017. In the report, Electronic Person the concept was put forward, and it was stated that a hybrid regulation should be adopted creating a sui generis framework for liability in situations arising from the autonomous behaviour of artificial intelligence, while in other situations it would be appropriate to proceed under the relevant legislation. This is because artificial intelligence is affected by its environment, adapts over time and behaves accordingly. This introduces unpredictability. The law does not concern itself with the unpredictable; it concerns itself with certain events.
In conclusion, artificial intelligence is not recognised as a legal subject in the current Turkish legal system. In this respect, there is no objection to treating it as an object, and liability must be imposed on the persons who hold the capacity of operator, creator and owner.
On the basis that artificial intelligence is an object, liability will arise under Turkish positive civil law in two respects: contractual and non-contractual liability. Under criminal law, since artificial intelligence has no personality, it may be regarded as a tool that facilitates the commission of an offence.
Contractual liability may be considered here in two ways. First, the subject matter of the contract in question may be the manufacture and sale of artificial intelligence tools. In this respect, the liability relationship does not arise from artificial intelligence; rather, recourse may be had to the provisions on warranty against defects in Article 219 et seq. TCO concerning the liability of the manufacturer, seller and buyer under the contract of sale. This situation does not arise from liability directly caused by artificial intelligence.
The matter with which we are principally concerned is the conclusion of a contract with artificial intelligence or the performance of a contract by means of artificial intelligence. Here, liability for damage arising in the conclusion or performance of the contract must be determined. Under our positive law, as set out in Article 112 et seq. TCO, liability for the non-performance or improper performance of a contractual obligation lies with the person using the artificial intelligence. For example, during the current pandemic, while an obligation is being established by means of software, a contract that should not have been concluded may be concluded, or it may be concluded differently from how it should have been, owing to an error originating from the software. In this respect, liability lies with the firm, company or similar entity using this software. Within this scope, the user may also be held strictly liable under Article 116 TCO on liability for the acts of auxiliary persons.
Damage may arise as a result of the autonomous behaviour of artificial intelligence and operations originating from its software. Who will be liable for conduct arising from an error in the software that causes such damage? It is more appropriate to address this issue in terms of fault-based liability and, under strict liability, liability for failure to exercise due care and liability for hazardous activities.
Considering first liability for failure to exercise due care, the provisions on this form of liability in our positive law are found in Article 66 et seq. TCO. Various separate situations are listed there. Since the law sets them out by way of enumeration, the cases of strict liability are exhaustive. There is no case of strict liability relating to artificial intelligence in the law. Therefore, no non-contractual liability arises on the basis of liability for failure to exercise due care, a form of strict liability, for conduct caused by artificial intelligence.
Some views argue that the exhaustively listed cases of strict liability should be applied to artificial intelligence by analogy. If we examine them one by one, it is very easy to see that on some very fundamental points this is not possible in certain cases.
Applicability by analogy of strict liability cases to artificial intelligence
| Provision | Sorumluluk hali | Assessment in the article |
|---|---|---|
| TBK. md. 66 | Employer’s liability | Not possible, since artificial intelligence is not a natural person and no employment relationship can be established |
| TBK. md. 68 | Liability of the owner of a building | Application by analogy is impossible, since artificial intelligence cannot be a construction attached to the ground |
| TBK. md. 67 | Liability of the keeper of an animal | For artificial intelligence, this is the most suitable case for application by analogy |
Article 66 TCO “liability of the employer” concerns the employer’s liability; here, artificial intelligence would first have to be a natural person, and an employment relationship would also have to be established between it and the employer. Neither of these is possible.
Article 68 TCO “liability of the owner of a building” concerns the building owner’s liability; according to the settled case law of the Court of Cassation, a “building” must be a “constructed work attached to the ground”. Since it is not possible for artificial intelligence to be a constructed work attached to the ground, application by analogy is also impossible under this article.
Article 67 TCO “liability of the keeper of an animal” concerns the animal keeper’s liability and is the provision most suitable for application by analogy to artificial intelligence. The first paragraph of the article provides: “a person who permanently or temporarily undertakes the care and management of an animal is liable to compensate for the damage caused by the animal.” In this respect, it may be considered that artificial intelligence should likewise give rise to liability under this provision for damage resulting from its autonomous behaviour. Although artificial intelligence is expected to possess a more advanced level of intelligence than such animals, it takes decisions when displaying autonomous behaviour and, in some cases, may take these decisions on its own volitional initiative, as animals do, and may also act outside orders and instructions. In this respect, the situations bearing the greatest similarity arise here.
“a person who permanently or temporarily undertakes the care and control of an animal is obliged to compensate for the damage caused by the animal.”
Turkish Code of Obligations, Art. 67/1
Pozitif hukukta durum
These are the views on artificial intelligence from the perspective of liability for failure to exercise due care, which we have now examined. It is worth bearing in mind that no strict liability relationship has yet been established for artificial intelligence in our positive law.
Liability for hazardous activities is regulated in Article 71 TCO, which establishes a form of strict liability based on situations creating a general hazard. The first part of the article establishing general liability describes liability for hazardous activities as follows: “Where damage arises from the activity of a business posing a significant hazard, the owner of the business and, if any, the operator shall be jointly and severally liable for that damage.” The first paragraph specifies who is to be liable and refers to a degree of hazard, namely “significant hazard”, which is defined in the second paragraph as follows: “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. In particular, where any statute provides for special hazard liability for businesses posing similar hazards, that business shall also be deemed to be a business posing a significant hazard.” This is how it is set out.
In line with the definitions in the provision above, a conclusion may be reached under the general provisions on liability for hazardous activities where damage arises in the operation of such enterprises as a result of a hazard caused directly by artificial intelligence used in part of their operations, or as a result of experts acting in breach of the duty of care and diligence required in the use of this artificial intelligence. The provisions on fault-based liability are, of course, reserved. However, where a special form of liability for hazardous activities applies, liability will be determined within the framework of its own legislation.
Fault-based liability, or in other words liability for tort, is addressed in Article 49 et seq. TCO. Article 49 provides that a person who causes damage through culpable and unlawful conduct is liable for that damage, and its second paragraph provides that, even if there is no legal rule prohibiting the conduct, the person will be liable for the damage if the culpable conduct is contrary to morality.
In this case, the burden of proving the damage and the harmful conduct lies with the injured party. If the amount of the damage suffered cannot be determined precisely, the judge determines it equitably, taking into account the ordinary course of events. The injured party is entitled to claim compensation. The special circumstances in determining compensation and the principles for calculating the amount to be awarded are regulated in Article 51 et seq. TCO.
In the case of damage caused by artificial intelligence, it is not possible to examine the fault of the artificial intelligence, because it cannot be held liable for its culpable conduct. The position here will vary depending on how the artificial intelligence is used.
To give an example: who will be liable for the damage resulting from an accident involving a driverless car (smart car)? Let us assume that the accident was caused by the driverless car. The other party injured in the accident may recover their damage from the owner of the vehicle, the operator of the vehicle and the insurance company. In an accident caused by a driverless car, the cause of the accident may stem from one or more of software or manufacturing errors. As a result, the parties liable to pay compensation will have recourse for the amount paid according to the degree of fault.
Liability according to the source of the error in the driverless car example
| Source of the error | Sorumlu |
|---|---|
| Software | Software company (employer’s liability); the company may take recourse internally according to the degree of fault |
| Manufacturing (mechanical components etc.) | Manufacturer |
| Both together | Joint and several liability; recourse in the internal relationship |
Many legal situations arise as a result of artificial intelligence permeating our lives. The points we have addressed in our study have been considered by various views within the framework drawn by our positive law, and the position is not clear-cut. Since law is a discipline that must renew itself, it is essential that legislation be developed on these legal situations and issues while we are still at the very beginning.
The article defines artificial intelligence as a form of reasoning performed by machines and information systems, and divides it into three levels: narrow artificial intelligence programmed for a single function, strong artificial intelligence capable of behaving like a human, and super artificial intelligence capable of human creativity and judgement.
The article gives as examples an autonomous vehicle being involved in an accident, software drafting contracts on behalf of a company making an error regarding essential elements, and data-collection software accessing personal data it should not access.
The article assesses the thing theory and the legal personality view, and argues that artificial intelligence is not recognised as a legal subject in the current Turkish legal system and that, being treated as an object, liability should be placed on the persons acting as its director, creator and owner.
Makaleye göre Avrupa Parlamentosu Hukuk İşleri Komisyonu 2017 yılında yayımladığı raporda ‘elektronik kişi’ kavramını ortaya atmış; yapay zekanın özerk davranışlarından doğan durumlarda sorumluluk bakımından kendine özgü bir alan yaratılmasını önermiştir.
According to the article, in this case liability for non-performance or improper performance of the obligation under TBK Art. 112 et seq. lies with the company using the artificial intelligence; the user may also be held liable for the acts of an auxiliary person under TBK Art. 116.
The article states that some views advocate this, and contends that the liability of employers and of building owners cannot be applied by analogy, whereas the liability of animal keepers is the most suitable case. According to the article, positive law contains no form of strict liability created for artificial intelligence.
According to the example in the article, the injured party may claim compensation from the vehicle's owner, operator and insurance company; those who pay have a right of recourse in proportion to the degree of fault, depending on whether the error stems from the software, from manufacturing, or from both.
Consultation request
At UTU Hukuk Danışmanlık, we provide legal representation and advisory services in the field of general law.
Fill in the form and we will call you. You can also contact us directly: