Artificial Intelligence (AI)-Generated Synthetic Evidence And The Criminal Trial In India: Rethinking Authentication, Proof, And Statutory Silence Under The Bharatiya Sakshya Adhiniyam, 2023
- IJLLR Journal
- Aug 4
- 2 min read
Dr. Priyadarshi Nagda, Assistant Professor, University College of Law, Mohanlal Sukhadia University, Udaipur
Zibran Khan Pathan, Research Scholar, University College of Law, Mohanlal Sukhadia University, Udaipur.
ABSTRACT
Generative Artificial Intelligence (GenAI) has quietly upended the assumptions on which criminal evidence law was built. Systems capable of producing convincing synthetic audio, video, still images and written documents grouped here under the umbrella term "AI-generated evidence" have opened a jurisprudential fault line that legal systems worldwide, India included, have yet to close. India's newest evidence statute, the Bharatiya Sakshya Adhiniyam, 2023 ("BSA"), which supplanted the one-hundred-and- fifty-year-old Indian Evidence Act, 1872 with effect from 1 July 2024, is in many respects a genuine modernisation of the country's proof regime. Yet, as this paper contends, that modernisation stops short precisely where it is needed most: the BSA offers no meaningful answer to deepfakes, machine- fabricated documents, algorithmic behavioural predictions, or the wider category of machine-learning output now entering criminal courtrooms.
Working through doctrinal analysis, comparative law and constitutional reasoning, the paper isolates five structural gaps in the BSA: (i) no definitional scaffolding exists for AI-generated evidence as a distinct category; (ii) the certificate-driven authentication regime in Section 63 was never designed to test synthetic content and cannot do so reliably; (iii) the so-called "liar's dividend" erodes confidence even in genuine digital evidence; (iv) chain-of-custody norms remain silent on evidence that has passed through, or been produced by, an AI system; and (v) no competency benchmark exists for judges or forensic examiners tasked with evaluating AI-based outputs. The paper places these gaps against the backdrop of three foreign models the still-developing US Federal Rule of Evidence 707, the European Union's Artificial Intelligence Act of 2024, and the United Kingdom's Forensic Science Regulator Act, 2021 to sketch out a reform agenda for India. It closes by proposing a new Schedule V-A to the BSA as a workable statutory template for regulating AI-generated evidence in Indian criminal trials.
