ABSTRACT

Deepfakes, the hyper-realistic, AI Generated synthetic media has precipitated an existential crisis for the administration of criminal justice globally and within India. For over a century, the Indian legal system operated on the visual epistemic assumption that “seeing is believing,” treating video recordings, CCTV footage, and photographs as highly cogent, objective pieces of documentary proof. However, AI can now synthesise human likenesses and voices with near perfect fidelity, rendering this assumption obsolete. This paper analyses the vulnerabilities of digital evidence under the new criminal law architecture of India, specifically focusing on the Bharatiya Sakshya Adhiniyam, 2023. Examining the intersection of Section 63 of the BSA, the Information Technology Act, 2000, and the Bharatiya Nyaya Sanhita, 2023 (BNS), the paper identifies critical gaps in the new dual-certification regime (Part A and Part B of the Schedule). It further evaluates the constitutional vulnerabilities under Articles 19, 20(3), and 21 of the Constitution of India, particularly highlighting the emergence of the “liar’s dividend.” Finally, it offers concrete, systemic recommendations for law enforcement, forensic laboratories, and the judiciary to re-engineer a robust “visual chain of custody” utilising cryptographic watermarking, mandatory blockchain-based hashing at the point of ingestion, and specialised judicial training.

THE EXORDIUM

The foundational architecture of the Indian law of evidence was drafted in the late nineteenth century, an era dominated by tangible, physical items. As technology advanced, the judiciary incrementally expanded its interpretation of documents to accommodate tape recordings, magnetic disks, and eventually digital outputs. Throughout this evolution, a stable jurisprudential thread persisted: while oral testimony could be tainted by human malice, bias, or memory decay, amachine generated record was presumed to capture an immutable slice of reality.

The advent of Deepfakes, synthetic audio, video, or images generated through artificial intelligence that convincingly depict individuals saying or doing things they never did, has fundamentally shattered this paradigm. Built on Generative Adversarial Networks (GANs), where two neural networks (a generator and a discriminator) compete to create and refine flawless fabrications, Deepfakes have shifted digital manipulation from crude pixel editing to sophisticated semantic deception. This technological shift introduces an unprecedented procedural crisis. The threat is two fold:

The Ingestion of Fabricated Proof: Innocent individuals can be falsely implicated in severe criminal offenses, ranging from financial fraud under the Bharatiya Nyaya Sanhita, 2023 (BNS) to offenses against the state, via synthetic digital video files that look in distinguishable from real recordings.

The “Liar’s Dividend”: Conversely, a guilty defendant can weaponise the known existence of Deepfake technology to introduce reasonable doubt, asserting that a perfectly authentic, incriminating video recording is merely an AI Generated fake.

This erosion of truth undermines the “visual chain of custody”, the unbroken chronological record documenting the collection, custody, transfer, and analysis of digital video or audio material. As India transitions into its newly overhauled criminal code framework, evaluating whether our evidentiary statutes can withstand this epistemic shock is not an academic exercise; it is an urgent structural necessity.