Published on: 26th July 2026
Authored by: Tanmay Deshmukh
Bharti Vidyapeeth, Pune
Abstract
Artificial intelligence (AI) is changing our lives in myriad ways—from healthcare to transportation and decision-making.[1] The proliferation of these systems, however, raises significant legal questions: Who is accountable when an AI system causes harm?[2] The doctrines of negligence and product liability—both premised on a world governed by human action and decision—have thus been tested by a generation of intelligent machines.[3] This paper examines the adequacy of India’s current laws to address harms caused by AI systems and contrasts them with analogous approaches in the European Union, the United States, the United Kingdom, and China.[4] Ultimately, it finds that India must embrace forward-thinking legislation to ensure robust liability frameworks that foster AI innovation while prioritizing transparency, accountability, and consumer protection.[5]
Keywords: Artificial Intelligence, Liability, Negligence, Product Liability, AI Regulation, Consumer Protection, Autonomous Systems, India.
Introduction
In an era dominated by technology, Artificial Intelligence (AI)—powered systems that simulate human cognitive functions—have permeated every facet of our lives.[6] AI technologies are being increasingly deployed in sectors as diverse as healthcare, transportation, finance, education, and law enforcement.[7] India, driven by a vision of enhanced innovation and economic development, is rapidly adopting AI.[8] Simultaneously, concerns regarding AI-driven risks such as accidents, discriminatory practices, and privacy violations are escalating.[9]
Traditional legal doctrines, designed for a human-centric legal environment, grapple with the inherent nature of AI.[10] These frameworks primarily rely on established concepts of fault, intent, and causation which often fail to align with the autonomous decision-making, learning capabilities, and distributed nature of AI systems.[11] If an autonomous vehicle injures a pedestrian, is the AI algorithm itself to blame, its developers, its manufacturers, or its owner?[12] This paper seeks to assess whether existing Indian legal principles—in particular tort law—are adequately equipped to handle such complexities and posits that India needs an effective framework tailored to address the unique challenges posed by AI.[13]
Understanding Artificial Intelligence and Autonomous Systems
Artificial Intelligence generally refers to the ability of a machine or a computer system to imitate intelligent human behavior, typically by processing large amounts of data and learning from it.[14] AI systems can analyze information, make predictions or decisions, and perform tasks that normally require human intelligence, such as problem-solving, speech recognition, and natural language processing.[15] Three broad categories of AI are generally recognized:
A. Narrow Artificial Intelligence: Narrow AI, also known as weak AI, is designed for a specific task and cannot perform beyond its pre-programmed capabilities. Examples include virtual assistants like Siri or medical diagnostic software.
B. General Artificial Intelligence: General AI, or strong AI, represents a more advanced form of intelligence capable of understanding, learning, and applying knowledge to any intellectual task a human being can. This form of AI has yet to be developed.
C. Autonomous Systems: Autonomous systems are machines that are capable of operating and making decisions with little or no direct human control. These include self-driving vehicles, robotic arms in manufacturing, and unmanned aerial vehicles.
The legal significance of AI technologies—particularly autonomous systems—stems from four key features:
1. Autonomy: Unlike humans, AI systems make autonomous decisions based on programming and data. This fundamentally challenges traditional notions of free will and responsibility in legal contexts.
2. Learning Capability: Most modern AI systems employ machine learning, meaning they evolve and improve over time as they are exposed to new data. This continuous learning means the behaviour of such systems is not static or perfectly predictable, creating uncertainty in liability assignment.
3. Opacity and the “Black Box” Problem: Many AI systems, particularly those using deep neural networks, are opaque in their functioning. It can be difficult or impossible to trace the specific inputs and algorithms that led to a particular decision. This lack of transparency is known as the “black-box” problem and poses significant hurdles in establishing causality and fault.
4. Distributed Responsibility: AI systems often involve multiple actors: the developers of algorithms, the manufacturers of hardware, providers of training data, and the users or operators of the AI system. Determining who is responsible when such systems fail due to faulty design, flawed data, or user error can be incredibly complex.
Traditional Liability Doctrines
Legal systems traditionally attribute responsibility for harm through various doctrines. The following doctrines are most relevant to AI-related liabilities:
1. Negligence: This common law tort requires proving a duty of care owed to the plaintiff, a breach of that duty, a causal link between the breach and the harm, and resultant damage. In AI cases, proving negligence involves demonstrating foreseeability, identification of the responsible party (who owed the duty), and the specific acts or omissions constituting a breach.
2. Product Liability: This doctrine holds manufacturers and sellers strictly liable for injury caused by a defective product, irrespective of whether the manufacturer exercised due care. The defect can be in design, manufacture, or warning. The application to AI is complicated by whether an AI algorithm can be considered a “product” and how to define and prove a “defect” in code that evolves through machine learning.
3. Vicarious Liability: This doctrine holds one party legally responsible for the tortious acts of another, usually where there is a relationship such as employer-employee or principal-agent. It is potentially applicable where an AI system is controlled and operated by an employee within their course of employment.
Why Traditional Liability Doctrines Fail in AI Cases
Traditional liability regimes are ill-equipped to deal with AI because they were conceived in a fundamentally different legal context—one where the agent of harm was consistently a human acting intentionally or negligently, whose thought processes and decision-making were accessible. AI’s characteristics radically disrupt these assumptions:
1. The Black-Box Problem: The inability to “look inside” an AI and understand the reasoning behind its actions is one of the greatest obstacles to applying negligence or product liability. Proving that a specific algorithm exhibited a “duty” or a “breach” becomes challenging when its internal workings are inscrutable. Courts find it difficult to infer fault and causation without clear explanations for the AI’s behavior, especially when multiple factors are in play.
2. Foreseeability Challenges: Negligence law hinges on the foreseeability of harm. AI, particularly when it uses machine learning, can develop capabilities or behave in ways not anticipated by its original creators. The ability of AI systems to learn and adapt, even in unintended ways, may expand the scope of foreseeable harm far beyond what was originally contemplated by designers or regulators, placing an undue burden on developers.
3. Multiple Stakeholders and Distributed Responsibility: When an AI system causes harm, the line of responsibility often becomes blurred among multiple parties. Consider a collision caused by an autonomous vehicle. Was it the AI’s software, a malfunction in the sensors, a failure in the vehicle’s control system, the cloud-based operating system the owner neglected to update, or the road design or environmental factors that contributed to the incident? Assigning fault within this complex network of interconnected actors—developers, hardware manufacturers, data providers, owners, regulators—is significantly more complicated than tracing fault in a typical human-made mistake. Traditional single-actor causality frameworks don’t easily accommodate this diffuse structure.
4. Causation Difficulties: Establishing causation—a direct link between the defendant’s conduct and the plaintiff’s injury—is a cornerstone of tort law. The dynamic and complex nature of AI, coupled with issues of opacity, learning, multiple stakeholders, and unpredictable external factors, makes proving definitive causation a formidable task. For instance, how does one definitively prove whether a diagnostic error from an AI system stemmed from an issue in its initial design, a lack of adequate training data, improper usage by the medical practitioner, or simply the inherent complexity and inherent error rate of the condition itself?
5. The Question of AI Legal Personhood: As AI capabilities advance, some legal scholars have begun exploring the possibility of granting AI legal personality—perhaps limited—to attribute legal rights and duties to AI systems themselves. In theory, an AI could be made responsible for its own actions, similar to how corporations are held responsible. However, this is a highly controversial idea, as AI systems currently lack sentience, moral consciousness, independent wealth, or the other characteristics typically associated with legal persons. Most legal academics find AI legal personhood to be an inadequate solution to the liability challenges, as it sidesteps the core issue of how humans—designers, manufacturers, operators—should be held accountable for the actions of autonomous systems.
Comparative Analysis: Global Approaches to AI Liability
Numerous jurisdictions are in the process of developing or have already begun to implement regulations and legal frameworks to address AI liability issues, often guided by a variety of philosophies.[16]
1. European Union: The European Union has taken a proactive and comprehensive approach to AI regulation, placing a strong emphasis on risk management and a human-centric approach.
* AI Act: The EU’s flagship legislation, the Artificial Intelligence Act (AI Act), categorizes AI systems based on their potential risk level: unacceptable-risk, high-risk, limited-risk, and minimal-risk. High-risk systems face stringent obligations concerning data governance, transparency, documentation, and human oversight.
* AI Liability Directive: Proposed to complement the AI Act, this directive aims to facilitate compensation for individuals suffering damage caused by AI systems by easing the burden of proof on victims.
2. United States: The United States has largely adopted a decentralized, sector-specific approach to regulating AI rather than a singular, overarching statute.
* Sectoral Regulations: Federal agencies such as the FCC and NHTSA are establishing guidelines and regulations for AI use within their respective domains.
* Product Liability and Negligence Doctrines: American courts adjudicate AI liability disputes on a case-by-case basis, applying existing tort and product liability principles.
3. United Kingdom: The UK government’s AI strategy emphasizes a principles-based and iterative approach to regulation rather than establishing a single regulatory body.
* Principles-Based Regulation: Focuses on broad principles of responsible AI use—including safety, fairness, transparency, and accountability—enforced by existing regulatory bodies across sectors.
4. China: China has emerged with a swift and comprehensive regulatory approach characterized by state oversight and social stability.
* Specific AI Applications & Algorithmic Transparency: Implements detailed regulations for recommendation systems, deepfakes, and algorithmic transparency, mandating impact assessments and security checks.
The varied global approaches to AI liability offer valuable lessons for India.[17] India can draw on these models to formulate its own strategy, one that balances the need for effective regulation with the imperative to encourage innovation.[18]
The Indian Legal Framework
Currently, India does not have a dedicated statute explicitly governing liability for AI systems.[19] Existing legal principles, particularly those found in tort law, along with certain provisions in consumer protection and information technology legislation, offer partial recourse.[20]
Information Technology Act, 2000: The Information Technology Act is India’s primary piece of legislation related to cyber-security and e-commerce. Although conceived and enacted before the advent of modern AI technologies, the Act contains provisions related to liability for data breaches, electronic records, and security incidents. The IT Act, with its framework for intermediary liability, may serve as a starting point. However, it does not specifically address issues like the opacity of AI algorithms or distributed liability arising from complex AI systems.
References
[1] Donoghue v. Stevenson, [1932] AC 562 (HL).
[2] Rylands v. Fletcher, (1868) LR 3 HL 330.
[3] M.C. Mehta v. Union of India, (1987) 1 SCC 395.
[4] Consumer Protection Act, No. 35 of 2019, India Code (2019).
[5] Information Technology Act, No. 21 of 2000, India Code (2000).
[6] Digital Personal Data Protection Act, No. 22 of 2023, India Code (2023).
[7] Ministry of Electronics and Information Technology, National Strategy for Artificial Intelligence, Government of India (2018).
[8] OECD, Recommendation of the Council on Artificial Intelligence, OECD/LEGAL/0449 (2019).
[9] European Parliament and Council, Artificial Intelligence Act, Regulation (EU) 2024/1689.
[10] European Commission, Proposal for a Directive on Adapting Non-Contractual Civil Liability Rules to Artificial Intelligence (AI Liability Directive), COM (2022) 496 final.
[11] John McCarthy, What Is Artificial Intelligence?, Stanford University (2007).
[12] Stuart Russell & Peter Norvig, Artificial Intelligence: A Modern Approach (4th ed. 2020).
[13] Frank Pasquale, The Black Box Society: The Secret Algorithms That Control Money and Information (2015).
[14] Mireille Hildebrandt, Law for Computer Scientists and Other Folk (2020).
[15] Woodrow Barfield & Ugo Pagallo, Research Handbook on the Law of Artificial Intelligence (2018).
[16] Ryan Calo, Robotics and the Lessons of Cyberlaw, 103 Calif. L. Rev. 513 (2015).
[17] European Union Agency for Fundamental Rights, Getting the Future Right: Artificial Intelligence and Fundamental Rights (2020).
[18] NITI Aayog, Responsible AI for All: Approach Document for India (2021).
[19] Alan Turing, Computing Machinery and Intelligence, 59 Mind 433 (1950).
[20] Luciano Floridi & Josh Cowls, A Unified Framework of Five Principles for AI in Society, 1 Harv. Data Sci. Rev. (2019).
