THE LIABILITY DEBATE: WHO PAYS WHEN AI ERRORS?

A Comprehensive Legal Analysis of Artificial Intelligence & Tort Liability in India

Author: Gopikrishna P R | Institution: ABBS SCHOOL OF LAW, Karnataka State Law University

1. Introduction: The Rise of Autonomous Systems in India Artificial Intelligence (AI) has transitioned from an experimental software tool to a fundamental driver of decision-making across key economic sectors in India. Automated credit-scoring algorithms determine financial eligibility, machine-learning diagnostics assist medical professionals in identifying oncological anomalies, algorithmic recruitment tools screen job applicants, and computer-vision models support autonomous navigation. While these technological advancements offer unprecedented efficiency, scalability, and economic value, they also introduce unprecedented operational and legal vulnerabilities. When a human professional—such as a physician, loan officer, or driver—commits an error resulting in physical harm, financial loss, or unlawful discrimination, classical jurisprudence provides well-established remedies through the law of torts, contract law, and statutory liability. However, when an autonomous or semi-autonomous AI system generates a harmful error, the traditional frameworks of fault, duty of care, and proximate causation face severe conceptual limitations. The central legal inquiry confronting Indian legal scholars, policymakers, and litigants is both clear and complex: Who pays when AI errors?

2. The Tripartite Liability Dilemma: Developer, Deployer, or Operator?

Allocating legal responsibility for AI-generated harm is inherently complicated because the AI lifecycle spans multiple distinct entities, each exerting varying degrees of control over the system's execution:

1. The AI Developer: The software vendor, algorithm architect, or technology firm that codes, trains, and validates the core machine-learning model. Should developers be strictly liable for structural flaws, data bias, or algorithmic hallucinations that emerge long after deployment?

2. The Commercial Deployer: The enterprise, medical institution, or financial firm that integrates the AI tool into its operational workflow. Does relying on an automated decision-making tool shift the primary burden of care onto the organization benefiting from its deployment?

3. The End-User Operator: The individual professional (e.g., a doctor or loan officer) who acts upon the AI's output. Does an operator bear sole liability for failing to independently verify an algorithmic recommendation, or is it unreasonable to expect human intervention when evaluating opaque software?

The 'Black Box' Barrier: Machine learning models, particularly deep neural networks, function through non-linear mathematical weights that are opaque even to their original developers. This opacity severely impedes a claimant's ability to satisfy the traditional tort standard of proving specific operational negligence or breach of duty

3. Statutory Evaluation: Current Indian Legal Landscape India currently lacks a dedicated statute governing civil and tort liability for artificial intelligence. Consequently, legal actions involving AI failure must rely on a fragmented framework of existing statutes and common law doctrine

Information Technology (IT) Act, 2000

Section 79 grants 'Safe Harbor' protection to passive intermediaries. Developers frequently claim intermediary immunity for third-party automated outputs. However, because Section 79 was enacted for static platforms, it is ill-equipped to address active generative AI models.  

Consumer Protection Act, 2019

Chapter VI introduces statutory 'Product Liability' for defective goods, allowing consumers to sue manufacturers or service providers for damages caused by defects. Nevertheless, ambiguity arises because software is frequently classified legally as a service rather than a tangible product.  

DPDP Act, 2023

Sections 6 to 10 establish Data Fiduciary duties and obligations specifically for algorithmic processing, regulating lawful processing, bias mitigation, and personal data rights. However, its primary focus remains strictly on personal data privacy, omitting civil tort compensation for financial or physical AI harms.  

Law of Torts (Common Law)

This framework relies on traditional doctrines such as Negligence, Strict Liability (Rylands v. Fletcher), and Vicarious Liability. While it offers judicial flexibility to hold deep-pocket deployers or creators accountable, proving proximate cause and foreseeability in complex deep-learning systems remains virtually impossible.  

4. Practical Case Studies in the Indian Context

Case Study A: Algorithmic Discrimination in FinTech Credit Allocation

Consider a prominent Indian FinTech lender utilizing an automated credit-scoring algorithm. If the model relies on historical training data containing implicit socio-economic biases, it may systematically deny credit or impose higher interest rates on applicants from specific demographic regions. When a rejected applicant challenges this decision, the lender may claim it merely relied on a third-party software vendor, while the vendor asserts that the deployment data was controlled by the bank. Under current Indian law, proving intentional discrimination or negligence requires deconstructing propriety algorithmic models, leaving the consumer without an effective remedy.

Case Study B: Diagnostic Failure in Health-Tech Systems

In a diagnostic imaging center, an AI radiology tool fails to detect an early-stage malignant tumor, leading to delayed treatment and aggravated patient injury. The attending radiologist relied on the software's 'clear' diagnostic report. If the patient files a medical negligence suit under the Consumer Protection Act, courts face a severe dilemma: Should the radiologist be held vicariously liable for relying on state-of-the-art diagnostic software, or should product liability attach to the Health-Tech vendor that developed the diagnostic AI?

5. The Way Forward: Proposed Regulatory & Legal Frameworks

To resolve the existing legislative vacuum and foster responsible AI innovation in India, legal scholars and policy analysts propose a multi-tiered regulatory framework:

A. Risk-Based Liability Categorization: Statutory frameworks should classify AI applications by risk tier. High-risk deployment (e.g., medical diagnostics, credit scoring, autonomous transportation) should be subject to strict liability standards, whereas low-risk applications (e.g., recommendation engines) remain governed by general negligence.

B. Mandatory AI Risk Insurance Pools: Commercial deployers of high-risk AI tools should be required by statute to maintain specialized third-party liability insurance. Similar to motor vehicle insurance, this guarantees prompt financial compensation for injured claimants without requiring complex evidentiary proof of algorithmic fault.

C. Statutory Human-in-the-Loop (HITL) Mandates: Legal liability should strictly attach to commercial deployers who execute high-stakes automated decisions without meaningful human oversight. Requiring documented human validation ensures a identifiable legal entity remains accountable for final operational outcomes.

6. Conclusion & Reader Discussion

As India positions itself as a global leader in digital infrastructure and technology adoption, establishing clear, predictable legal rules for AI liability is paramount. Balancing consumer protection with technological innovation requires moving beyond outdated legislative paradigms and enacting specialized, risk-proportionate AI legislation.

Join the Legal Discussion:

Who do you believe should bear primary civil liability when an autonomous AI system causes financial or physical harm—the software developer, the commercial deployer, or the end-user? Share your legal insights and comments below!

AUTHOR PROFILE & ACADEMIC CREDENTIALS

Author Name: Gopikrishna P R

Institution: ABBS SCHOOL OF LAW, Karnataka State Law University

Area of Specialization: Student of Law specializing in Artificial Intelligence Law, Technology Policy, Cyber Law, and Intellectual Property Rights in India.

Academic Focus: Researching civil liability frameworks, algorithmic transparency, and data governance in automated decision systems.