AI-Generated Evidence and the Indian Criminal Process: Challenges of Reliability and Admissibility

Abstract Artificial intelligence has transformed the evidentiary environment of criminal justice. Audio, video, images, documents and text can now be generated or altered with a degree of realism that makes ordinary visual or auditory inspection increasingly unreliable. India’s new evidence framework, the Bharatiya Sakshya Adhiniyam, 2023 (BSA), recognises electronic and digital records while imposing specific safeguards concerning their source, integrity, certification and hash values. The Supreme Court’s decision in Pune Bar Association v Union of India (2026) has upheld the constitutional validity of Section 63(4) and expressly recognised the heightened risks created by artificial intelligence and deepfakes. This article argues that admissibility, authenticity, reliability and probative weight must nevertheless remain analytically distinct. A technically compliant certificate can establish important aspects of provenance and integrity without establishing that an AI-generated representation is truthful, historically accurate or attributable to the person against whom it is tendered. The article examines synthetic media, AI enhancement, voice cloning, generated documents, algorithmic identification and machine-assisted forensic analysis. It analyses Sections 39, 61, 62 and 63 of the BSA, the relationship between the new regime and the Supreme Court’s Section 65B jurisprudence, and the fair-trial implications of opaque or insufficiently testable AI processes. It proposes a five-stage AI Evidence Reliability and Admissibility Test based on provenance, integrity, methodological reliability, contextual attribution and independent corroboration. The article concludes that Indian courts should neither exclude digital evidence merely because AI was involved nor presume that technical authenticity establishes truth. A calibrated, process-sensitive approach is required to preserve both technological utility and the presumption of innocence.

Authors

Institutions

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-10-08
DOI
https://doi.org/10.5281/zenodo.23233478
Primary Topic
Education, Law, and Society
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

AI-Generated Evidence and the Indian Criminal Process: Challenges of Reliability and Admissibility

Srishti Punj
Zenodo (CERN European Organization for Nuclear Research)
Education, Law, and Society
article

AI-Generated Evidence and the Indian Criminal Process: Challenges of Reliability and Admissibility

Srishti Punj
article en

Abstract

Abstract Artificial intelligence has transformed the evidentiary environment of criminal justice. Audio, video, images, documents and text can now be generated or altered with a degree of realism that makes ordinary visual or auditory inspection increasingly unreliable. India’s new evidence framework, the Bharatiya Sakshya Adhiniyam, 2023 (BSA), recognises electronic and digital records while imposing specific safeguards concerning their source, integrity, certification and hash values. The Supreme Court’s decision in Pune Bar Association v Union of India (2026) has upheld the constitutional validity of Section 63(4) and expressly recognised the heightened risks created by artificial intelligence and deepfakes. This article argues that admissibility, authenticity, reliability and probative weight must nevertheless remain analytically distinct. A technically compliant certificate can establish important aspects of provenance and integrity without establishing that an AI-generated representation is truthful, historically accurate or attributable to the person against whom it is tendered. The article examines synthetic media, AI enhancement, voice cloning, generated documents, algorithmic identification and machine-assisted forensic analysis. It analyses Sections 39, 61, 62 and 63 of the BSA, the relationship between the new regime and the Supreme Court’s Section 65B jurisprudence, and the fair-trial implications of opaque or insufficiently testable AI processes. It proposes a five-stage AI Evidence Reliability and Admissibility Test based on provenance, integrity, methodological reliability, contextual attribution and independent corroboration. The article concludes that Indian courts should neither exclude digital evidence merely because AI was involved nor presume that technical authenticity establishes truth. A calibrated, process-sensitive approach is required to preserve both technological utility and the presumption of innocence.

Zenodo (CERN European Organization for Nuclear Research)
Panjab University (IN)
Openalex Percentile: Top 3%
Education, Law, and Society
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

AI-Generated Evidence and the Indian Criminal Process: Challenges of Reliability and Admissibility — Srishti Punj · Zenodo (CERN European Organization for Nuclear Research) (2026) | TGRS Research Map | TGRS