Synthetic Media and the Law of Personal Identity in India: Deepfakes, Platform Liability and the Constitutional Limits of the Information Technology Amendment Rules, 2026

Abstract

Generative artificial intelligence has made the fabrication of a person’s face, voice and gestures inexpensive, rapid and convincing. India responded on 10 February 2026 by notifying the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Amendment Rules, 2026, which for the first time define “synthetically generated information”, require prominent labelling and embedded provenance metadata, oblige significant social media intermediaries to obtain and technically verify user declarations about synthetic content, and compress the removal deadline for flagged unlawful material to three hours. This paper evaluates that framework against the constitutional standards governing intermediary liability and freedom of expression, and against the private law of personality that Indian courts have been constructing case by case since 2022. It argues that Indian law is now organised around the wrong regulatory object. The 2026 Rules treat synthetic media principally as a problem of deception, to be cured by disclosure; but the gravest harms — non-consensual intimate imagery, sexual humiliation, identity-based fraud and dignitary injury — are not cured by a label, because they are not diminished by the audience knowing that the depiction is fabricated. Simultaneously, the compressed timelines, the incentive to over-remove, the reliance on automated detection and the absence of independent review sit uneasily with the reading down of section 79(3)(b) of the Information Technology Act, 2000 in Shreya Singhal, with the proportionality standard in K.S. Puttaswamy, and with the Bombay High Court’s reasoning in Kunal Kamra. The paper further shows that the judicial remedy that presently works — the wide John Doe injunction protecting personality rights — is available in practice only to claimants of celebrity, leaving ordinary victims with a criminal process ill-adapted to speed and anonymity. Drawing on the EU AI Act, China’s labelling measures and the United States TAKE IT DOWN Act, the paper proposes a statutory right in one’s likeness and voice, a harmdifferentiated remedial structure, evidentiary reform under the Bharatiya Sakshya Adhiniyam, 2023, and independent oversight of takedown power.