Author- Mr. Anuj Sethi
Research Scholar, Dr. Bhimrao Ambedkar University, Agra
Email: sethianuj488@gmail.com
Co- Author: Vidwanshu Chauhan
Student, City Academy Law College, University of Lucknow,
Email: vidwanshuchauhan@gmail.com
ABSTRACT
The rapid development of generative artificial intelligence (“AI”) has transformed the evidentiary risks associated with digital information. Photographs, videos, audio recordings, and documents can now be generated or manipulated with a degree of realism that increasingly challenges conventional methods of authentication. This development creates a significant difficulty under the Bharatiya Sakshya Adhiniyam, 2023 (“BSA”): a digital file may be technically intact while the representation contained within it is entirely artificial.
The BSA establishes a statutory framework for electronic and digital records through Sections 61–63, including certification requirements and technological safeguards such as hash values.¹ These safeguards are essential to establishing the integrity and provenance of an electronic record. They do not, however, necessarily establish whether the underlying representation accurately depicts a real-world event. This article therefore distinguishes between digital integrity and representational authenticity.
Drawing upon the Supreme Court’s recent decision in Pune Bar Association v. Union of India and the broader judicial concern surrounding verification of AI-generated material, this article proposes a FiveLayer Authentication Framework consisting of provenance, digital integrity, forensic content analysis, contextual corroboration, and judicial evaluation. The framework seeks to reconcile technological safeguards with traditional principles of evidence while preventing courts from treating technical authenticity as synonymous with substantive truth.
Keywords: Artificial Intelligence; Deepfakes; Digital Evidence; Bharatiya Sakshya Adhiniyam; Electronic Records; Authentication; Hash Value; Expert Evidence
I. INTRODUCTION
The courtroom has entered an evidentiary environment in which seeing is no longer necessarily believing. Artificial intelligence systems can generate realistic photographs, imitate voices, produce convincing videos, and alter existing recordings. Deepfake technology has therefore transformed the central question[1] surrounding digital evidence from merely “Has this file been altered?” to the more difficult question: “Does this representation correspond to reality?”[2]
This distinction is particularly important under the Bharatiya Sakshya Adhiniyam, 2023 (“BSA”), which replaced the Indian Evidence Act, 1872 and governs evidence in proceedings to which it applies from July 1, 2024.² Sections 61–63 establish the principal statutory framework for electronic and digital records.³ Section 63 additionally prescribes conditions concerning the admissibility of electronic records, including certification and technological information relating to the source and integrity of the record.⁴ These safeguards are necessary. Yet a significant conceptual gap remains.
CORE PROPOSITION
An authentic digital file can contain an inauthentic representation.
A perfectly preserved deepfake may possess an unbroken hash value. A genuine device may contain a manipulated recording. A properly certified electronic record may still depict an event that never occurred.
Accordingly, authentication of the file should not automatically be equated with authentication of the representation.
II. THE BSA FRAMEWORK FOR ELECTRONIC EVIDENCE
The BSA expressly recognises electronic and digital records as evidence and provides a statutory mechanism governing their admissibility. Sections 61–63 are central to this framework.⁵
Section 63 establishes conditions for the admissibility of electronic records. Its accompanying Schedule requires information concerning the electronic device or source and includes the hash value of the relevant electronic record together with the algorithm used to generate that value.⁶
Hashing is important because a hash value can operate as a digital fingerprint. If a file changes, its hash will ordinarily change as well. This provides the court with an important mechanism for testing whether the electronic record presented is substantially the same as the record identified at the relevant stage.
However, hashing answers a limited question:
Is the digital object substantially the same as the object from which the hash was generated?
It does not answer:
Is the content of that object a truthful representation of a real-world event?
This distinction becomes critical in the context of AI-generated evidence.
III. DEEPFAKES AND THE NEW AUTHENTICATION PROBLEM
Traditional approaches to digital evidence have largely focused on the possibility that a record was altered after its creation. Deepfakes complicate that assumption because manipulation may occur at the moment of creation itself.
Consider a hypothetical video showing a public official making a statement. If the original file is preserved and its hash remains unchanged, the electronic record may satisfy conventional integrity requirements. But if the video was generated through AI rather than recorded from an actual event, its technical integrity does not establish its factual authenticity.
The evidentiary problem therefore contains at least four distinct dimensions:
- Admissibility
Can the material legally be received by the court?
- Digital Integrity
Has the electronic file been altered after acquisition or preservation?
- Representational Authenticity
Does the content accurately represent the event, person, voice, image, or communication it purports to show?
- Probative Value
Even if authentic, what weight should the court assign to it?
These questions should not be collapsed into a single concept of “authentication.”
IV. PUNE BAR ASSOCIATION: HASHING, CERTIFICATION AND THE LIMITS OF AUTHENTICATION
The Supreme Court’s recent consideration of these issues in Pune Bar Association v. Union of India provides an important development in the law of electronic evidence.⁷
⁷ 7. Pune Bar Ass’n v. Union of India, W.P. © No. 599 of 2026, order (S.C. May 22, 2026).
The petition challenged, among other matters, the requirements under Section 63(4) and its Schedule. The Supreme Court recognised the special difficulty associated with electronic records because digital information may be continuously modified, copied, transferred, or manipulated. The Court also recognised that AI and deepfake technology intensify concerns regarding authenticity, integrity, and probative value.⁸
The Court observed that the requirement of a hash value had a rational connection with establishing the authenticity and integrity of electronic records. It further recognised the value of expert certification in providing additional assurance concerning the technical characteristics of the record.⁹
Importantly, however, the Court did not conclusively determine the broader constitutional challenge to Section 63. The decision should therefore not be understood as holding that every electronic record satisfying the statutory certification requirements is substantively authentic.
The Court also considered the interaction between Section 39 of the BSA and Section 79A of the Information Technology Act, 2000. Section 39 recognises expert opinion, including persons possessing special skill in a relevant field.¹⁰ The Court indicated that, where supported by unimpeachable material, a person possessing specialised knowledge in computer science or cyber forensics may potentially qualify as an expert for purposes of the statutory framework.
The Court further clarified that an earlier view requiring certification by a Section 79A Examiner should not be treated as binding precedent.
The significance of Pune Bar Association, therefore, lies not merely in hashing or certification. It demonstrates that digital evidence requires a more sophisticated conception of authenticity.
V. FROM ELECTRONIC-RECORD AUTHENTICITY TO REPRESENTATIONAL RELIABILITY
Indian electronic-evidence jurisprudence has historically focused on establishing the reliability of electronic records.
In Anvar P.V. v. P.K. Basheer, the Supreme Court established important principles concerning electronic records and the statutory requirements applicable under the former Indian Evidence Act.¹¹ These principles were subsequently clarified in Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal.¹²
Those decisions remain important historical foundations, but the technological environment has changed significantly.
The BSA must now operate in a world where an electronic record can be generated without a corresponding real-world event. Thus, a future approach to authentication should distinguish between three concepts:
⁸Id. ⁹Id. ¹⁰ Bharatiya Sakshya Adhiniyam, No. 47 of 2023, § 39(1)–(2) (India); Information Technology
Act, No. 21 of 2000, § 79A (India). ¹¹ Anvar P.V. v. P.K. Basheer, (2014) 10 SCC 473 (India). ¹²Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, (2020) 7 SCC 1 (India).A. Record Authenticity
Whether the electronic object presented is what it purports to be.
B. Representational Authenticity
Whether the content accurately depicts or communicates what it purports to depict or communicate.
C. Evidentiary Reliability
Whether the circumstances surrounding the creation, preservation, transmission, and use of the evidence justify reliance upon it.
The second and third questions become particularly important for deepfakes.
VI. THE FIVE-LAYER AUTHENTICATION FRAMEWORK
To address this emerging problem, this article proposes a Five-Layer Authentication Framework for AI-sensitive digital evidence.
① PROVENANCE
Where did the evidence originate?
↓
② DIGITAL INTEGRITY
Has the file remained unchanged?
↓
③ FORENSIC CONTENT ANALYSIS
Is there evidence of synthetic generation or manipulation?
↓
④ CONTEXTUAL CORROBORATION
Does independent evidence support the representation?
↓
⑤ JUDICIAL EVALUATION
What evidentiary weight should the court assign?
1. Provenance
The first layer concerns the origin of the material. Courts should examine who created or captured the recording, which device was used, how the material was obtained, and how it reached the investigating agency or litigant.
A recording of uncertain origin should receive greater scrutiny than evidence whose chain of custody is clearly established.
2. Digital Integrity
The second layer corresponds most closely with the BSA’s existing technological safeguards. Hash values, metadata, acquisition procedures, storage protocols, and certificates can help determine whether the digital object has remained unchanged.
This is necessary but insufficient.
3. Forensic Content Analysis
The third layer addresses whether the representation itself appears synthetic or manipulated. Depending upon the nature of the evidence, examination may consider facial movement, audio characteristics, metadata anomalies, compression patterns, lighting, shadows, frame transitions, or other forensic indicators.
Expert evidence becomes especially important at this stage.
Section 39 of the BSA permits opinions from persons specially skilled in relevant fields, while Section 39(2) specifically refers to the Examiner of Electronic Evidence under Section 79A of the Information Technology Act.¹³
4. Contextual Corroboration
No technical examination should operate in isolation. Courts should examine whether independent evidence supports the alleged event.
For example, a video may be assessed alongside witness testimony, contemporaneous messages, location records, other recordings, documents, or surrounding circumstances.
The objective is not to demand corroboration in every case but to recognise that synthetic media can defeat purely file-based authentication.
¹³Bharatiya Sakshya Adhiniyam, No. 47 of 2023, § 39(1)–(2) (India); Information Technology Act, No. 21 of 2000, § 79A (India).
5. Judicial Evaluation
The final layer belongs to the court.
Even evidence that passes all technical tests does not automatically become conclusive proof. The court must determine its relevance, reliability, consistency with other evidence, and ultimate probative value.
The framework therefore preserves the traditional judicial function while adapting authentication to AI generated evidence.
VII. EXPERT EVIDENCE AND FORENSIC EXAMINATION
The increasing sophistication of synthetic media makes expert evidence increasingly significant.
However, courts should not treat the word “AI” as a substitute for methodology. An expert opinion should identify the material examined, the methodology used, the limitations of the examination, and the basis for the conclusion.
This is particularly important because AI-detection tools themselves are not infallible. A court should therefore avoid treating a single automated detection score as conclusive proof that a recording is genuine or fake.
Instead, forensic conclusions should ideally be evaluated alongside provenance, integrity, contextual evidence, and other relevant material.
The BSA’s expert-evidence provisions provide a statutory foundation for such an approach. The challenge is to ensure that expert evidence assists the court without transferring the judicial function of determining truth to a technological system.
VIII. THE BROADER 2026 JUDICIAL APPROACH TO AI VERIFICATION
The Supreme Court’s decision in Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd. illustrates a broader judicial concern with verification in the age of AI.¹⁴
The case did not concern deepfake evidence directly. Instead, the Court dealt with judicial reliance upon non-existent or AI-generated legal authorities. The decision nevertheless demonstrates an important principle: AI-assisted processes cannot eliminate the responsibility of human verification.
¹⁴ Pooja Ramesh Singh v. Jammu & Kashmir Bank Ltd., 2026 INSC 668, Civil Appeal No. 11950 of 2025 (India July 2, 2026).
This principle has direct relevance to digital evidence.3
If AI can generate false legal authorities convincingly, it can also generate false audiovisual representations convincingly. In both situations, the solution is not to reject technology altogether but to introduce stronger verification mechanisms.
Human judgment must therefore remain central.
IX. PROCEDURAL AND CONSTITUTIONAL BALANCE
A robust authentication system must balance two competing concerns.
On one hand, weak authentication can allow fabricated evidence to influence judicial decisions. On the other hand, excessively burdensome procedures may prevent genuine electronic evidence from being used effectively.
The BSA should therefore be interpreted in a manner that distinguishes admissibility safeguards from substantive evaluation.
Section 63 certification should establish the necessary technological foundation for receiving electronic evidence. It should not be treated as a statutory declaration that the underlying representation is truthful.
3 Similarly, forensic examination should assist the court rather than create an absolute technological test for authenticity.
A flexible, evidence-sensitive approach is preferable because technology changes faster than legislation.
X. RECOMMENDATIONS
First, courts should expressly distinguish digital integrity from representational authenticity when dealing with AI-sensitive evidence.
Second, Section 63 certificates should clearly establish what they certify: the source, manner of production, integrity, and relevant technological characteristics of the record.
Third, courts should encourage appropriate forensic examination where the circumstances raise a reasonable possibility of synthetic manipulation.
Fourth, expert reports should disclose methodology and limitations rather than merely state a conclusion.
Fifth, judicial officers should evaluate digital evidence holistically, including provenance, technical integrity, forensic findings, and independent corroboration.
Finally, institutional guidelines should be developed for the preservation and examination of AI-sensitive digital evidence so that investigators, lawyers, experts, and courts apply reasonably consistent procedures.
XI. CONCLUSION
The emergence of deepfakes exposes a fundamental limitation in conventional approaches to electronic evidence: a file can be genuine while its representation is false.
The Bharatiya Sakshya Adhiniyam, 2023 provides important technological safeguards through Sections 61–63, including certification and hash-based integrity mechanisms. Yet these safeguards cannot, by themselves, establish that an image, video, or audio recording accurately represents reality.
The Supreme Court’s discussion in Pune Bar Association demonstrates the importance of technological integrity, expert examination, and the changing risks posed by AI-generated material. At the same time, the broader judicial approach to AI verification reinforces the continuing importance of human scrutiny.
The future of digital evidence law should therefore move beyond a binary question of whether a file is “authentic.” Courts should instead ask a sequence of questions:
Where did it come from? Has it been altered?
Was the representation synthetically generated?
Is it corroborated?
What evidentiary weight should it receive?
The proposed Five-Layer Authentication Framework offers one possible structure for answering these questions.
The central principle is simple:
Digital integrity is evidence of the integrity of a file—not necessarily evidence of the truth of the story that the file tells.
As generative AI becomes increasingly capable of manufacturing convincing representations of reality, preserving this distinction will be essential to the integrity of adjudication.
BLUEBOOK FOOTNOTES
- Bharatiya Sakshya Adhiniyam, No. 47 of 2023, §§ 61–63 (India).
- Id.; Bharatiya Sakshya Adhiniyam, No. 47 of 2023, commencement provision (India).
- Bharatiya Sakshya Adhiniyam, No. 47 of 2023, §§ 61–63 (India).
- Id. § 63(4) & sched.
- Id. §§ 61–63.
- Id. § 63(4) & sched.
- Pune Bar Ass’n v. Union of India, W.P. (C) No. 599 of 2026, order (S.C. May 22, 2026).
- Id.
- Id.
- Bharatiya Sakshya Adhiniyam, No. 47 of 2023, § 39(1)–(2) (India); Information Technology Act, No. 21 of 2000, § 79A (India).
- Anvar P.V. v. P.K. Basheer, (2014) 10 SCC 473 (India).
- Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, (2020) 7 SCC 1 (India).
- Bharatiya Sakshya Adhiniyam, No. 47 of 2023, § 39(1)–(2) (India); Information Technology Act, No. 21 of 2000, § 79A (India).
Pooja Ramesh Singh v. Jammu & Kashmir
[1] Bharatiya Sakshya Adhiniyam, No. 47 of 2023, §§ 61–63 (India).
[2] Id.; Bharatiya Sakshya Adhiniyam, No. 47 of 2023, commencement provision (India). ³ Bharatiya Sakshya Adhiniyam, No. 47 of 2023, §§ 61–63 (India).
⁴ Id. § 63(4) & sched. ⁵ Id. §§ 61–63.
⁶ Id. § 63(4) & sched.
