THE DEEPFAKE DILEMMA: A TIERED JUDICIAL STANDARD FOR AUTHENTICATING AI-GENERATED EVIDENCE UNDER PAKISTANI LAW

Abstract

The advent of generative artificial intelligence has now allowed the creation of photographic, video, and audio evidence of a quality that can be easily challenged, posing a critical threat to the courts in Pakistan and other countries. When a court declines disputed audio-visual content without sufficient justification, a real litigant or victim will have been deprived of a lawful remedy; when it permits disputed content to pass without sufficient verification, a court may be making a legal decision or conviction based on a fake. The paper discusses how the current evidentiary architecture in Pakistan, which is mainly the Qanun-e-Shahadat Order 1984, the Electronic Transactions Ordinance 2002 and the Prevention of Electronic Crimes Act 2016, as amended in 2025, remains relevant but is incomplete for wholly synthetic evidence, and the aim is to address this predicament. The paper, based on statutory text, reported case law, and existing Pakistani legal scholarship holds that provenance, chain of custody, integrity, and non-tampering considerations used in Pakistan are incomplete to deal with the wholesale synthetic fabrication, as they were originally created in a pre-generative-AI context. It suggests a more graded judicial authentication criterion - a threshold proponent burden, a burden-shifting burden on true challenge, institutionalized judicial gatekeeping, and an institutional forensic backstop - designed to minimize and reduce erroneous exclusion and erroneous reliance without necessitating wholesale new law.