WHEN SEEING IS NO LONGER BELIEVING: DEEPFAKES, ARTIFICIAL INTELLIGENCE AND THE CRISIS OF DIGITAL EVIDENCE IN INDIA

Abstract

Generative artificial intelligence has unsettled the evidentiary status of digital media. Audio, images and video can now be generated or manipulated with a plausibility that no longer reliably signals factual authenticity, and the resulting harms range across impersonation, fraud, reputational injury, privacy violation, non-consensual intimate imagery, political manipulation and the tendering of fabricated material in legal proceedings. This article asks whether India's cyber, criminal, evidentiary and constitutional framework is equipped to meet that challenge, and argues that the deeper problem is not the arrival of a new category of cybercrime but a broader crisis of digital authenticity. It draws two distinctions that existing scholarship has tended to collapse: file integrity from event authenticity, and attribution from accountability. Building on these distinctions, it proposes an original A³D Framework — Authenticity, Attribution, Accountability and Due Process — together with a Five-Question Deepfake Evidence Test addressing source, integrity, authenticity, attribution and corroboration. It further proposes a Digital Authenticity Protocol for India resting on provenance, preservation, forensic verification, judicial assessment and calibrated accountability. A comparative examination of the European Union's transparency-oriented model illustrates the value of provenance and disclosure while underscoring the continuing need for constitutional safeguards. The article concludes that Indian law should adopt a layered analytical sequence

— admissibility, integrity, authenticity, reliability and probative weight — in place of the binary “real or fake” determination that current practice still tends to assume.