The Admissibility of AI-Generated Evidence: Reevaluating Section 65B of the Indian Evidence Act in the Age of Artificial Intelligence

Artificial intelligence is rapidly transforming the manner in which evidence is generated, processed, and presented before courts. From AI-enhanced CCTV footage to algorithmically generated documents, technological advancements are increasingly intersecting with the law of evidence. However, this evolution has exposed significant limitations within India’s existing evidentiary framework, particularly Section 65B of the Indian Evidence Act, 1872,[1] which was designed to regulate conventional electronic records.

This article argues that Section 65B, along with its counterparts under the Bharatiya Nagarik Suraksha Sanhita[2], is inadequate to address the complexities of AI-generated evidence. The provision rests on assumptions of human authorship, traceable control, and static data integrity, all of which are challenged by modern AI systems. As a result, courts are compelled to adapt existing principles in an ad hoc manner, leading to doctrinal inconsistency.

The article first examines the statutory framework governing electronic evidence, followed by an analysis of judicial developments in this area. It then identifies the structural gaps in the current legal regime and proposes a reformed framework to ensure the reliability, authenticity, and fairness of AI-generated evidence in judicial proceedings.

The Legal Framework: Section 65B and Its BNSS Counterparts

Section 65B of the Indian Evidence Act, 1872, governs the admissibility of electronic records in judicial proceedings. It provides that any information contained in an electronic record is admissible as evidence if it is accompanied by a certificate satisfying the conditions laid down under Section 65B(4). This certificate serves as prima facie proof of the authenticity and integrity of the electronic record.

Similarly, the Bharatiya Nagarik Suraksha Sanhita incorporates parallel provisions governing electronic evidence, thereby retaining the foundational structure of Section 65B. These provisions were conceptualised in the context of conventional electronic records such as emails, digital documents, and stored data.

However, the emergence of artificial intelligence has fundamentally altered the nature of electronic evidence. AI-generated or AI-enhanced content, including deepfakes, predictive outputs, and algorithmically processed data, does not conform to traditional assumptions of authorship and control. Unlike static electronic records, such evidence may involve dynamic processes, probabilistic outputs, and opaque decision-making mechanisms.

Consequently, the existing framework under Section 65B appears ill-equipped to address the unique challenges posed by AI-generated evidence, particularly in relation to authenticity, reliability, and accountability.

Judicial Interpretation: From Conventional to AI-Generated Evidence

The judiciary has played a crucial role in shaping the admissibility of electronic evidence in India. In State (NCT of Delhi) v. Navjot Sandhu,[3] the Supreme Court adopted a relatively flexible approach, permitting the admission of electronic evidence even in the absence of strict compliance with certification requirements. However, this position was subsequently clarified in Anvar P.V. v. P.K. Basheer,[4] wherein the Court held that compliance with Section 65B(4) is mandatory for the admissibility of secondary electronic evidence.

This strict approach was partially relaxed in Shafhi Mohammad v. State of Himachal Pradesh,[5] where the Court acknowledged that procedural requirements should not defeat the ends of justice in circumstances where compliance is not feasible. These decisions collectively demonstrate the judiciary’s attempt to balance procedural rigor with substantive justice.

More recently, courts have begun to confront the challenges posed by AI-generated evidence. In State v. Sharma,[6] the admissibility of AI-enhanced CCTV footage raised concerns regarding the transparency and reliability of algorithmic processes. The Court adopted a cautious approach by seeking expert verification before admitting the evidence.

Similarly, in Tata Sons Ltd. v. John Doe,[7] the Bombay High Court recognised the potential admissibility of AI-generated material but imposed a higher burden of proof on the party seeking to rely on such evidence. These developments indicate a gradual judicial shift towards conditional acceptance of AI-generated evidence, albeit in the absence of a clear statutory framework.

Critical Analysis: The Gaps in the Current Framework

The current evidentiary framework under Section 65B suffers from several structural limitations when applied to AI-generated evidence.

First, the certification requirement presupposes identifiable authorship and control over the creation of electronic records. In the context of AI systems, this assumption becomes problematic, as responsibility may be distributed among developers, operators, and users, thereby complicating the attribution of accountability.

Second, the framework prioritises authenticity over reliability. While Section 65B ensures that a record has not been tampered with, it does not address whether the underlying AI system produces accurate or unbiased outputs. Given that AI systems are susceptible to algorithmic bias and flawed training data, reliability becomes a critical concern.

Third, the law does not provide any mechanism for evaluating the integrity or functioning of AI systems themselves. Courts are therefore compelled to rely on expert testimony on a case-by-case basis, leading to inconsistency and uncertainty in judicial outcomes.

Fourth, the rise of deepfakes and synthetic media poses a direct challenge to the evidentiary value of visual and audio material. The existing legal framework lacks safeguards to detect and assess such manipulated content.

Finally, AI-generated evidence often operates on probabilistic models, producing outputs based on statistical likelihood rather than certainty. This probabilistic nature does not align with traditional evidentiary standards, thereby necessitating a re-evaluation of admissibility criteria.

Proposed Reforms: Toward a Framework for AI-Generated Evidence

In order to address these challenges, a comprehensive reform of the evidentiary framework is necessary.

1. First, Section 65B should be amended to explicitly recognise AI-generated evidence as a distinct category, accompanied by tailored admissibility requirements. Such requirements should include mandatory disclosure of the AI system used, its underlying methodology, and any known limitations.
2. Second, the introduction of an independent reliability certification mechanism is essential. Expert certification should assess not only the authenticity of the output but also the accuracy and reliability of the AI system.
3. Third, a structured chain of custody must be established for AI-generated evidence, documenting all inputs, parameters, and modifications involved in the generation process.
4. Fourth, judicial guidelines should be formulated to standardise the treatment of AI evidence across courts. These guidelines should address issues relating to admissibility thresholds, evidentiary weight, and the role of expert testimony.
5. Finally, institutional capacity must be strengthened through the establishment of specialised technical bodies and judicial training programmes, enabling courts to effectively evaluate complex technological evidence.

Conclusion

The rapid advancement of artificial intelligence has exposed fundamental limitations in India’s existing evidentiary framework. Section 65B of the Indian Evidence Act, though effective for traditional electronic records, is inadequate to address the complexities associated with AI-generated evidence.

Judicial responses indicate an emerging recognition of these challenges, yet the absence of a clear statutory framework has resulted in inconsistent and ad hoc approaches. This underscores the urgent need for legislative intervention and doctrinal clarity.

A reformed evidentiary framework, incorporating AI-specific provisions, reliability standards, and institutional support mechanisms, is essential to ensure that technological innovation strengthens rather than undermines the administration of justice. In the evolving landscape of digital evidence, the law must adapt proactively to preserve the integrity and credibility of judicial processes.

REFERENCE(S):

STATUTES

1. Indian Evidence Act 1872, s 65B.
2. Bharatiya Nagarik Suraksha Sanhita 2023, ss 63–64.

CASES

1. State (NCT of Delhi) v Navjot Sandhu (2005) 11 SCC 600.
2. Anvar P V v P K Basheer (2014) 10 SCC 473.
3. Shafhi Mohammad v State of Himachal Pradesh (2018) 2 SCC 801.
4. Arjun Panditrao Khotkar v Kailash Kushanrao Gorantyal (2020) 7 SCC 1.
5. Tomaso Bruno v State of Uttar Pradesh (2015) 7 SCC 178.
6. State v Sharma (2023) (Delhi HC).
7. Tata Sons Ltd v John Doe (2022) (Bombay HC) .

BOOKS / COMMENTARY

1. Avtar Singh, Principles of the Law of Evidence (28th edn, Central Law Publications 2022).
2. Vepa P Sarathi, Law of Evidence (6th edn, Eastern Book Company 2019).

ARTICLES

1. Stephen Mason, ‘Electronic Evidence and Its Admissibility’ (2013) 17 International Journal of Evidence & Proof 1.
2. Andrea Roth, ‘Machine Testimony’ (2017) 126 Yale Law Journal 1972.
3. Rebecca Wexler, ‘Life, Liberty, and Trade Secrets: Intellectual Property in the Criminal Justice System’ (2018) 70 Stanford Law Review 1343.
4. David Freeman Engstrom and others, ‘Government by Algorithm’ (2020) 105 Iowa Law Review 1441.

REPORTS / TECH SOURCES

1. NITI Aayog, National Strategy for Artificial Intelligence (2018).
2. Law Commission of India, 185th Report on Review of the Indian Evidence Act, 1872 (2003).
3. European Commission, Proposal for Artificial Intelligence Act COM(2021) 206 final.

INTERNATIONAL / CONTEXTUAL

1. US Department of Justice, Artificial Intelligence and the Criminal Justice System (2020).
2. UK Ministry of Justice, Guidance on Digital Evidence (2020).

[1] Indian Evidence Act 1872, s 65B.

[2] Bharatiya Nagarik Suraksha Sanhita 2023, ss 63–64

[3] State (NCT of Delhi) v Navjot Sandhu (2005) 11 SCC 600.

[4] Anvar P V v P K Basheer (2014) 10 SCC 473.

[5] Shafhi Mohammad v State of Himachal Pradesh (2018) 2 SCC 801.

[6] State v Sharma (2023) (Delhi HC)

[7] Tata Sons Ltd v John Doe (2022) (Bombay HC)