ARTIFICIAL INTELLIGENCE AND TORT LIABILITY

ABSTRACT

With its large presence in the modern world of today, Artificial Intelligence (AI) has transformed virtually all aspects of today’s world through the use of AI in various areas, such as healthcare, transportation, finance and law enforcement. Systems that operate by themselves – including things, such as self-driving cars, medical diagnostic software, and intelligent algorithms – are performing functions that would previously have required human discretion. While AI has the potential to increase efficiency and be innovative, there is also a great deal of legal concern when these systems cause harm or injure someone. The use of traditional tort laws (negligence, duty of care, causation, product liability) to evaluate disputes between disagreeing parties concerning the actions of an AI system has become increasingly difficult given the unpredictable nature and autonomous actions of technology. The nature of these types of decisions was created to regulate the conduct of humans but they also have significant limitations in terms of applying these definitions to the decision-making capabilities of machines.

This article explores how AI relates to tort liability and how AI is challenging today’s legal system. This article discusses issues like gaps in liability, difficulty proving negligence and problems with responsibility and causation. This article looks at how jurisdictions like the U.S., U.K., and India have dealt with these challenges and finds that tort law will provide a good base to work off, but will not give any valid answers about these issues against autonomous vehicles. The paper concludes with a call to create new legal structures suited for these independent vehicles by developing new laws that create an appropriate balance between new technology and victim protection/accountability.

KEYWORDS : Artificial Intelligence, Tort Liability, Negligence, Autonomous Systems, Product Liability, Legal Accountability, Causation, AI Regulation.

INTRODUCTION

Modern society relies heavily on Artificial Intelligence (AI) technology that is being implemented across numerous industries including Health Care, Transportation, Finance, and Communication. The emergence of self-driving vehicles, Virtual Assistants and automated decision-making processes has enabled machines to perform some of the same tasks that once required a human’s mental capacity. The increase in efficiency and creativity achieved through the application of AI will also raise substantial issues relating to liability when those machines injure or damage other people or property; these potentialities warrant further examination.[1]

As there is growing usage of autonomous technologies there are still many intricate ways to determine who will be liable in the event of an accident ordering such technology. For instance, the traditional tort concept of negligence, or duty of care (which traditionally focused on human behavior) will be an increasingly relevant tool for assessing whether to permit or deny liability to a manufacturer, programmer, owner/user, or others engaged in usage of artificial intelligence systems. AI may have an autonomous path of operation after a particular set of variables and a programmer is removed from the equation, leading to difficulty determining which person should be liable for damage from an accident involving autonomous AI technology.[2]

Donoghue v. Stevenson established a landmark case paving the way for modern negligence law[3], but applying these principles in cases involving AI-related harm can be difficult. This article discusses the relationship between tort liability for AI systems and provides an analysis of the shortcomings of traditional tort law when dealing with AI systems as well as pointing out the necessity for reform to provide accountability and victim protection in today’s digital world.

UNDERSTANDING ARTIFICIAL INTELLIGENCE AND TORT LAW

Artificial intelligence (AI) is the ability of machines or computer programmes to exhibit human-like behaviour through completing tasks that normally require cognitive ability (e.g., learning, reasoning, decision-making, and problem-solving). The way this occurs is through a combination of machine learning technology and algorithmic processing/analysis to derive or analyze data so that decisions can be made by AI with little to no involvement from humans. AI is currently used across many different industries (e.g., Healthcare, transportation system, financial services, security), making AI an integral part of society today.[4]

A tort refers to a damage done by one individual against another as a result of a wrongful act. The purpose of tort law is to remedy such injuries through providing compensation to injured parties and placing liability on those responsible for the injury. Key concepts within torts are duty of care, breach of duty, causation, and strict liability. The application of these concepts to modern torts was established in the landmark case of Donoghue v Stevenson.

Due to the autonomous nature of AI systems, the relevance of AI and tort law has risen significantly. Tort law has been developed to deal with human actions, but AI systems work autonomously, often unpredictably. This raises challenges when determining how to attribute responsibility for harm caused by an AI system. Courts and legal systems are facing difficulties in determining whether liability lies with the manufacturer, programmer, owner or user.[5]

CHALLENGES OF APPLYING TRADITIONAL TORT LAW TO ARTIFICIAL INTELLIGENCE

1. Problems with Determining Duty of Care

The primary obstacle to applying traditional tort law to AI relates to finding the proper party liable for all damages suffered by a victim as a result of an autonomous system causing injury (either personally or property damage). Traditional tort law allows for recovery of damages from negligent human actors as the proximate causes of harm. The complexity of both the multiple parties involved (e.g., manufacturers, programmers, software developers, owners/users) and the algorithms controlling the operation (software and hardware) of an AI system creates uncertainty as to what parties should be responsible for tortious liability.[6]

2. Challenges in Proving Negligence

Traditional negligence law relies upon the theory that humans commit faults and use reasonable foresight when performing actions. Machine-learning-based Artificial Intelligence (AI) systems can develop independent decision-making capabilities that sometimes differ from those of their creators and may not always be foreseeable. This makes it difficult to determine whether or not negligence has occurred, or whether or not reasonable care was exercised by the developer and operator of an AI system, because of the “black box” nature of AI.[7]

3. Causation Issues

One of the hardest parts of filing a tort claim against AI is causing damage. Under tort law, for a person to win his/her suit, he/she must prove that the negligent act of the defendant significantly contributed to the damages alleged incurred by the plaintiff. Many independent variables used by AI systems result in the system coming up with multiple ways to arrive at the same conclusion, making it difficult to isolate each of the multiple independent variables from one another for purposes of demonstrating a cause and effect relationship with respect to the one identifiable variable. A specific example would be that if someone gets into an accident with an autonomous vehicle (AV) and there was a software error, a mechanical failure, or some other type of failure involved, each of these could be the reason that the person was either injured or killed in an accident with the AV.[8]

4. Liability Gap and Accountability Concerns

Due to not having a legal identity as a separate entity, AI or autonomous systems can’t be directly sued when they cause harm. This can leave victims without an avenue to seek damages and will present problems for tort law principles regarding accountability and liability.[9]

5. Product Liability Concerns

Many of these AI technology issues will also be applicable to product liability law. Generally, product liability laws are focused on the defects associated with physical products; however, when you start looking at software, there will be other forms of defects that may be associated with the AI, such as algorithmic bias, software issues, or constantly learning issues. It becomes difficult to assess whether or not the AI product is defective because the product will change and continue to be modified or evolve once it has been put into the market.[10]

COMPARATIVE LEGAL APPROACHES

Countries across the globe have different methods of processing claims of bad behaviour created by Artificial Intelligence (AI), whether it be with existing negligence and product liability principles as is the case in the USA; or through the enactment of specific new laws and regulations to identify who may be liable for the outcome of using AI, or what constitutes and acceptable level of care as defined by statute.

The courts in the United States apply established rules of tort law when addressing disputes involving AI, this includes evaluating whether an AI manufacturer/developer/operator exercised reasonable care to prevent harm caused by their device using what would be considered to be an appropriate standard of conduct as determined by judicial analysis of available evidence. The courts will look primarily to existing doctrines in tort law for fortifying the imposition of vicarious liability, to establish duty or provide for foreseeable damages when there was insufficient evidence to establish independent negligence giving rise to the imposition of liability to the AI manufacturer/developer/operator.

While the European Union has taken a much stricter regulatory and consumer protection-oriented position toward the governance of AI, it has incorporated risk-based regulations into their legal framework so that developers and providers of high-risk AI technology are subject to greater obligations for transparency, accountability and safety in AI systems. This trend also seeks to protect individuals from potential physical harm that may arise from the use of AI.[11]

In contrast to the EU’s more comprehensive framework, India’s current laws do not yet provide an explicit definition for AI liability. Additionally, there are no specific laws in India to govern AI, so Indian courts will rely more on traditional tort law as well as the consumer protection act 2019 and IT law to determine liability for case involving technological harm.[12] However, as the use of AI continues to grow within India and other types of industries, the need for a comprehensive legal framework to define liability, provide accountability and establish compensation for victims will continue to increase.

NEED FOR LEGAL REFORMS

1. New Laws for AI Liability

There are currently insufficient laws dealing with the issue of liability that arises from the usage of autonomous technology (AI).[13]

2. Clearly Allocate Responsibility

Legal systems need to place a clear definition on the responsibility of the manufacturer, programmer, developer, owner, or user of an AI system.[14]

3. Impose Strict Liability on High Risk AI Systems

Artificial Intelligence systems classified as high risk (e.g., self-driving vehicles & medical AI) need to fall under strict liability rules to allow for compensation to the injured party.[15]

4. Require Transparency & Explainability of AI Systems

Artificial Intelligence systems are required to operate in a transparent manner so that the decisions made by an algorithm can be explained and/or audited.[16]

5. Require Insurance for Autonomously Operated Technologies

Use of insurance for autonomously operated technology will provide a means to compensate victims in the event of an accident or injury related to the failure of an AI-based technology.[17]

CONCLUSION

The emergence of automated technology in our current society has revolutionized everyday life through the use of autonomous machines and systems that can carry out functions that would typically demand human intellect. In addition to helping both private and governmental organizations achieve higher levels of productivity and innovation, there are numerous types of legal issues related to liability and accountability with regard to the use of these technologies. Since traditional tort law exists primarily to regulate conduct among human beings, the application of these principles , including duty of care, negligence and causation, to autonomous/self-learning technologies has exposed significant gaps in these principles.

The article speaks of some types of challenges in building a bridge between tort law and AI, such as duty of care determination, finding out whether there is negligence, showing cause and effect, and treating issues of liability. The damages due to AI are objected to law systems by the jurisdictions such as the United States, the European Union, and India, all lacking a cohesive regulatory system or unifying legislation.

By and large, the advancement of technology in AI requires multiple modifications in existing regimes of liability, transparency, and victims’ protection to ensure accountability in AI. Current statutory liability regimes do not provide sufficient guidance regarding the rising technology threats that emerging technologies, such as AI, pose. Determination of accountability will need to bring in liability laws, formulating comprehensive regulatory frameworks and ethical governance models particular to AIs. This is to ensure that along with innovations, the public interest may be protected regarding illnesses or damage caused by Artificial Intelligence technology.

REFERENCES / BIBLIOGRAPHY

Books

1. Artificial Intelligence: A Modern Approach, Stuart Russell & Peter Norvig, Artificial Intelligence: A Modern Approach (4th ed., Pearson, 2021).
2. The Laws of Robots, Ugo Pagallo, The Laws of Robots: Crimes, Contracts, and Torts (Springer, 2013).

Journal Articles

1. Andrew D. Selbst, ‘Negligence and AI Systems’ (2020) 100 Boston University Law Review.

Cases

1. Donoghue v Stevenson [1932] AC 562 (HL).
2. Rylands v Fletcher (1868) LR 3 HL 330.

  Legislations and Legal Documents

1. Consumer Protection Act, 2019.
2. European Union AI Act Proposal.
3. Restatement (Third) of Torts: Products Liability.
4. Information Technology Act, 2000.

[1] Stuart Russell & Peter Norvig, Artificial Intelligence: A Modern Approach (4th ed., 2021)

[2] Andrew D. Selbst, ‘Negligence and AI Systems’ (2020) 100 Boston University Law Review

[3]Donoghue v Stevenson [1932] AC 562 (HL)

[4] Stuart Russell & Peter Norvig, Artificial Intelligence: A Modern Approach (4th ed., 2021)

[5] Ugo Pagallo, The Laws of Robots (Springer, 2013)

[6] Andrew D. Selbst, ‘Negligence and AI Systems’ (2020) 100 Boston University Law Review

[7] Ugo Pagallo, The Laws of Robots (Springer, 2013)

[8] Andrew D. Selbst, ‘Negligence and AI Systems’ (2020) 100 Boston University Law Review

[9] Ugo Pagallo, The Laws of Robots (Springer, 2013)

[10] Restatement (Third) of Torts

[11]European Union AI Act Proposal

[12] Consumer Protection Act, 2019

[13] Ugo Pagallo, The Laws of Robots (Springer, 2013)

[14] Andrew D. Selbst, ‘Negligence and AI Systems’ (2020)

[15]Rylands v. fletcher (1868) LR 3 HL 330

[16] European Union AI Act Proposal

[17] Ugo Pagallo, The Laws of Robots (Springer, 2013)