Inefficiency of Existing Insurance Laws in Accidents Caused by the Independence of Artificial Intelligence

When artificial intelligence in a self-driving car acts contrary to the instructions provided by the manufacturer or software developers, without any fault on the part of the programmer or developer, the manufacturer cannot be held responsible unless it can be proven that the programmer should have foreseen the risks and failed to do so. For example, if a self-driving car is programmed to only identify white people, it may not recognize a black person as human and cause an accident.
However, when AI acts outside its instructional framework based on its deep (reinforcement) learning and goes out of control, determining liability becomes complex. Who is responsible for compensating the injured parties? Several solutions have been proposed, including granting the car an independent legal personality, the model proposed by Abraham and Robin regarding the responsibility of the manufacturing company, and the proposal of a no-fault fund. After examining all these ideas and identifying their shortcomings and ineffectiveness, a final solution will be introduced under the title of a special compensation fund for artificial intelligence.
Automobiles have always provided humanity with new conveniences while also introducing risks and accidents. Therefore, the method of compensating victims of these accidents is a crucial issue in civil liability and civil liability insurance. This issue requires competent support, as consumers of technology expect that before the introduction of new technologies, along with the preparation of road infrastructure and legal infrastructure, the legal framework should also be reviewed and adjusted. Self-driving cars will be subject to this scrutiny.
Iran's legal system lacks codified laws regarding the civil responsibility of artificial intelligence in deep learning. Consequently, we must rely on general laws in this area. Therefore, it is necessary for lawyers to examine this legal issue, which will soon become a significant legal challenge for our country, based on existing laws and propose legal solutions for amending or introducing new legislation.
Currently, semi-autonomous cars like the Tigo 8 are already driving on the country's roads, and the use of artificial intelligence extends beyond self-driving cars, encompassing fields like medicine, household appliances (such as smart robotic vacuum cleaners), and new technologies such as the intelligent robot developed by Pila Technology Company in Mashhad, capable of opening doors, turning on lights, and retrieving apples from the refrigerator. Furthermore, the widespread use of ChatGPTs and the Iranian smart robot Sorena, which was ranked among the top ten robots of the year, demonstrates the country's significant progress in robotics. However, if any of these robots malfunction, determining liability for compensation becomes problematic, especially when neither the manufacturer, designer, nor user is at fault. This underscores the necessity of this study.
This article, albeit insignificant, can pave the way for addressing existing and future problems, as it emphasizes the importance of proactive measures. Another factor that increases the necessity of such studies is the core issue of responsibility and insurance in self-driving cars, which has compelled thinkers in this field to explore alternative solutions.
The use of a car, whether self-driving or traditional, inherently introduces societal risks. Therefore, the owner must bear the responsibility of using it. Due to the high risk and potential for unreimbursed losses, compulsory insurance is mandated. If a violation occurs and is not covered by insurance, the fund assumes responsibility.
The problem in self-driving cars is analogous, but with a crucial difference: the distinction between self-driving and traditional cars lies in the realm of insurance. Self-driving cars present the challenge of deep learning, making it difficult to pinpoint the culprit of an accident and identify the responsible party for insurance. In traditional cars, a human factor is usually or most of the time at fault, and the insurance company of the at-fault party compensates for the damage. However, in self-driving cars, accidents caused by deep learning are not attributable to any of the participants. The self-driving car itself, through its acquired learning, causes the accident. As we will discuss further, artificial intelligence cannot be held independently responsible. Therefore, it remains unclear who is responsible for insurance and whether all cases of accidents caused by deep learning should be compensated by the government and a dedicated fund. This further emphasizes the necessity of conducting studies like this article.
Almost all scientific literature on self-driving vehicles accepts as a central premise that self-driving vehicles will be significantly safer than traditional vehicles. More appropriately, the "brains" that constitute self-driving vehicles exhibit behavior that is often incomprehensible and therefore cannot be characterized as defective. While deeply rejecting artificial intelligence, these cases suggest that judges will soon be required to consider (and recognize) artificial intelligence, machine learning, and the impact of the use of these technologies, and to find an appropriate legal framework for them.
For this reason, many researchers propose entirely new systems to handle the liability of self-driving cars, ranging from immunity to absolute liability in the form of compulsory insurance. There are reasons to doubt all these hypotheses. The desire to provide compensation for people injured in self-driving cars without hindering the advancement of this life-saving technology has led legal scholars to consider alternatives to traditional civil liability and product liability. Two of the most frequently mentioned options are no-fault insurance and victim compensation funds. Both have a track record and can be adapted to meet the specific needs of the self-driving car market.
So far, articles in this field have generally addressed accidents caused by self-driving cars, or in general, the deep learning of artificial intelligence and self-driving cars, without providing concrete solutions. They have either focused on foreign laws and reviewed existing Iranian laws or have primarily dealt with criminal responsibility. One of the novel aspects of this article is its in-depth and case-by-case analysis of this issue, utilizing available judicial opinions worldwide and offering practical suggestions suitable for Iran's legal system.
This article distinguishes itself by examining incidents caused by deep learning on a case-by-case basis and explaining the proposed solutions. Various viewpoints have been proposed in this field, including the independent character of artificial intelligence (the flaws of which will be discussed later), the proposal of an all-government fund (which has disadvantages, as stated below), and the proposal of social cooperation. However, these proposals have not adequately addressed the issue of artificial intelligence's independence. Therefore, this article will present a new and more comprehensive proposal, and its advantages and framework will be further examined.
This article employs the following methodology:

Examination of the Theory of Civil Liability of the Car Based on Its Independent Character:This section will analyze the theory of assigning civil liability to the car itself as an independent entity and subsequently identify the flaws of this approach.
Examination of the Theory of Fault-Based Liability for Artificial Intelligence in the Field of Autonomous Vehicles:This section will explore the theory of holding artificial intelligence accountable based on fault in the context of autonomous vehicles.
Examination of Liability Without Fault for Artificial Intelligence and the Determination of the Compensation Fund:This section will delve into the concept of liability without fault for artificial intelligence and discuss the establishment of a compensation fund.
Examination of the Theory of the Government Compensation Fund:This section will analyze the theory of a government-funded compensation fund for damages caused by artificial intelligence.
Examination of the Theory of Abraham and Robin (Theory of the Responsibility of the Manufacturing Company):This section will examine the Abraham and Robin theory regarding the responsibility of the manufacturing company, outlining its advantages and disadvantages.
Presentation and Analysis of the Proposed System: A Special Fund for Compensation for Damages Caused by Deep Learning and the Autonomy of Artificial Intelligence:This final section will introduce and discuss the proposed system, its advantages, and its framework.