THE BLACK BOX OF AI: WHO’S TO BLAME?
AUTHOR – PREKSHA JAYASWAL* & DR. SHEFALI RAIZADA**
* STUDENT AT AMITY UNIVERSITY NOIDA
** DIRECTOR AND JOINT HEAD OF AMITY LAW SCHOOL, NOIDA
BEST CITATION – PREKSHA JAYASWAL, THE BLACK BOX OF AI: WHO’S TO BLAME?, INDIAN JOURNAL OF LEGAL REVIEW (IJLR), 5 (5) OF 2025, PG. 901-905, APIS – 3920 – 0001 & ISSN – 2583-2344
Abstract
Artificial Intelligence (AI), quickly embraced, or incorporated, into decision-making processes that have changed industries, also creates significant legal challenges. Most significantly, trust in AI is complicated by the “black box” nature of AI, which leads to a consternation about whether a decision is being made and how decisions are being made when everyone involved may be in the dark. This paper examines the disadvantages of non-transparency or lack of transparence of AI, and the issue of trying to assign fault, when AI systems cause injury (or breach of contract). We will look at traditional and any new theories of liability, and compare the laws concerning AI in the European Union, United States and India. The paper concludes with recommendations on the legal and policy front with the aim of bolstering accountability in the use of AI systems, and reliability of AI systems, recognizing that we are attempting to fix, or fill, critical gaps in existing legal and quasi-legal frameworks., with unique complications, showing that fault cannot easily be assigned febrile, ever-evolving machine learning models. Therefore, we must emphasis on transparency, explainability and ethical safeguards, as we argue for forward looking, legally young infrastructure and policy that reasonably encourages innovation while fostering public trust and responsibility.[1]
Keywords: Artificial Intelligence (AI), Black Box Problem, Legal Liability, AI Accountability, AI Transparency, Machine Learning, Autonomous Systems, Fault Theories, Negligence and AI, Strict Liability, Vicarious Liability.
[1] Binns, R. (2018). On the Importance of Transparency in AI Systems. Journal of Artificial Intelligence, 1(2), 14-29.