Author(s): Dr. Gazala Sharif and Adv. Palima Bhadra
Paper Details: Volume 4, Issue 3
Citation: IJLSSS 4(3) 15
Page No: 150 – 175
ABSTRACT
The rapid integration of Artificial Intelligence (AI) into legal systems has fundamentally altered the nature, production, and evaluation of evidence. AI-generated evidence—ranging from facial recognition outputs and predictive analytics to deepfake detection and automated forensic analysis—poses unprecedented challenges to traditional evidentiary principles under Indian law. This paper critically examines the admissibility of AI-generated evidence within the framework of the Indian Evidence Act, 1872 and its modern successor, the Bharatiya Sakshya Adhiniyam, 2023.
The study explores whether AI-generated outputs can be classified as “electronic evidence,” “expert opinion,” or a novel evidentiary category requiring independent recognition. It analyzes key statutory provisions, including Sections 65A and 65B of the Indian Evidence Act and Section 39 of the Bharatiya Sakshya Adhiniyam, focusing on requirements such as authenticity, reliability, and procedural compliance. The research also evaluates judicial trends in India concerning electronic evidence, alongside emerging global standards on AI evidence admissibility.
Recent developments, including judicial concerns over fabricated AI-generated citations and deepfake misuse, demonstrate the urgent need for doctrinal clarity and regulatory safeguards. The paper argues that while existing evidentiary principles can partially accommodate AI-generated evidence, significant gaps remain—particularly regarding algorithmic transparency, explainability, and accountability.
Through doctrinal and comparative analysis, this research proposes a hybrid evidentiary framework that incorporates traditional safeguards with AI-specific standards such as auditability, reproducibility, and human oversight. It concludes that without legislative reform and judicial innovation, the unchecked use of AI-generated evidence risks undermining the fairness, reliability, and integrity of judicial proceedings in India.
(Keywords: Artificial Intelligence (AI) Evidence, Electronic Evidence, Admissibility of Evidence, Section 65B Certification, Algorithmic Transparency, Expert Evidence, Deepfakes and Digital Manipulation, Reliability and Authenticity, Indian Evidence Law, Judicial Evaluation of AI)
INTRODUCTION
The Indian legal system is increasingly confronted with technological transformations that challenge its foundational doctrines. Among these developments, Artificial Intelligence (AI) stands out as a disruptive force, particularly in the domain of evidentiary law. AI-generated evidence defined as information produced, processed, or interpreted by machine learning systems, which has begun to influence criminal investigations, civil disputes, and administrative adjudications.
Traditionally, Indian evidence law has relied on human testimony, documentary records, and expert opinions. However, the emergence of AI systems capable of generating predictive insights, analysing large datasets, and producing synthetic media raises complex questions regarding admissibility, authenticity, and evidentiary value. The Indian Evidence Act, 1872, though amended to include electronic evidence, was not designed to address the autonomous and opaque nature of AI systems.
Recent incidents highlight the urgency of this issue. Indian courts have expressed concern over reliance on AI-generated or unverifiable legal materials, particularly where such content undermines judicial integrity. Additionally, the proliferation of deepfake technologies has raised serious concerns about the manipulation of digital evidence and its potential misuse in legal proceedings.
This paper seeks to examine whether existing evidentiary principles, such as relevance, admissibility, and reliability are sufficient to govern AI-generated evidence, or whether new legal standards are required. It also evaluates whether AI outputs can be treated as expert opinion or whether they constitute a distinct evidentiary category.
By analyzing statutory provisions, judicial precedents, and comparative frameworks, this study aims to provide a comprehensive understanding of the legal challenges posed by AI-generated evidence and to propose a robust framework for its admissibility under Indian law.
CONCEPTUAL FRAMEWORK OF AI-GENERATED EVIDENCE
AI-generated evidence refers to outputs created or processed by artificial intelligence systems, including: Facial recognition matches, Predictive policing outputs, Automated forensic reports, Deepfake detection results.
Artificial Intelligence (AI)-generated evidence refers to any evidentiary material that is produced, processed, or analyzed through algorithmic systems rather than direct human intervention. This includes outputs such as facial recognition matches, predictive policing data, automated forensic reports, voice recognition analysis, and deepfake detection results. Unlike traditional forms of evidence, AI-generated outputs are derived from probabilistic models trained on large datasets, making them inherently different in nature and reliability.
NATURE OF AI-GENERATED EVIDENCE: KEY CHARACTERISTICS
The distinction between traditional and AI-generated evidence lies in three fundamental characteristics: opacity, autonomy, and probabilistic reasoning. First, AI systems often operate as “black boxes,” meaning that their internal decision-making processes are not easily explainable or interpretable. This lack of transparency raises significant concerns regarding the verifiability of evidence. Second, AI systems operate with a degree of autonomy, generating outputs without continuous human oversight. Third, the conclusions produced by AI systems are probabilistic rather than deterministic, meaning they indicate likelihood rather than certainty.
From a legal perspective, this raises the question of classification. AI-generated evidence does not neatly fit into existing evidentiary categories such as documentary evidence or expert opinion under Section 45 of the Indian Evidence Act, 1872. While expert evidence traditionally involves human expertise, AI systems lack legal personality and accountability, thereby complicating their treatment within the evidentiary framework.
The Supreme Court in State of Maharashtra v. Sukhdev Singh emphasized that expert evidence must be subject to scrutiny and cross-examination to ensure reliability.^1 However, applying this standard to AI systems becomes problematic because algorithms cannot be cross-examined in the traditional sense. This creates a doctrinal gap in evidentiary law.
Furthermore, the emergence of synthetic media, particularly deepfakes, has intensified concerns about authenticity. The Delhi High Court in Anil Kapoor v. Simply Life India (2023) recognized the risks posed by AI-generated content and emphasized the need to protect personality rights against digital manipulation.^2 Although not directly addressing evidentiary admissibility, the case highlights judicial awareness of AI’s impact on legal processes.
PROBABILISTIC, NOT DETERMINISTIC
AI-generated outputs are inherently probabilistic rather than deterministic. Traditional evidence, such as eyewitness testimony or documentary proof, is typically presented as establishing facts with a degree of certainty, subject to verification and cross-examination. In contrast, AI systems rely on statistical models and machine learning algorithms that produce results based on likelihoods and patterns identified in training data.
For instance, a facial recognition system may indicate a “95% match” between an accused individual and an image captured at a crime scene. This does not establish identity conclusively but merely reflects a statistical probability derived from algorithmic processing. Such probabilistic outputs may be influenced by factors such as data quality, model design, and environmental conditions.
This characteristic raises significant concerns within the legal framework, particularly in criminal proceedings where the standard of proof is “beyond reasonable doubt.” The reliance on probabilistic evidence risks lowering this threshold, potentially leading to erroneous conclusions. Courts must therefore exercise caution in assigning evidentiary weight to AI-generated outputs and ensure that they are corroborated by independent evidence.
OPAQUE (“BLACK BOX”), LACKING EXPLAINABILITY
A defining feature of many AI systems, particularly those based on deep learning, is their lack of transparency or explainability—commonly referred to as the “black box” problem. Unlike human experts, who can articulate the reasoning behind their conclusions, AI systems often produce outputs without providing a clear or comprehensible explanation of the underlying decision-making process.
This opacity poses a serious challenge to the adversarial legal system, which relies on cross-examination and reasoned analysis to test the credibility of evidence. If the logic behind an AI-generated output cannot be scrutinized, it becomes difficult for courts to assess its reliability or for opposing parties to challenge its validity.
Moreover, the absence of explainability undermines judicial accountability. Judges are required to provide reasoned decisions based on evidence presented before them. Reliance on opaque AI outputs may compromise this requirement, as the reasoning process remains inaccessible. This issue has broader implications for procedural fairness and the right to a fair trial, particularly under constitutional principles.
DATA-DEPENDENT, RAISING BIAS CONCERNS
AI systems are fundamentally dependent on the data used for their training and operation. The quality, diversity, and representativeness of this data directly influence the accuracy and fairness of AI-generated outputs. If the training data contains biases—whether due to historical inequalities, sampling errors, or systemic discrimination—the AI system is likely to replicate and even amplify these biases.
For example, facial recognition systems have been shown to exhibit higher error rates for certain demographic groups, particularly women and individuals with darker skin tones. In a legal context, such biases can result in discriminatory outcomes, thereby violating principles of equality and non-discrimination.
The data-dependent nature of AI also raises concerns about accountability and transparency. In many cases, the datasets used for training AI systems are proprietary and not accessible for scrutiny. This lack of transparency makes it difficult to identify and rectify biases, further undermining the reliability of AI-generated evidence.
From a legal standpoint, the use of biased AI evidence may conflict with constitutional guarantees of equality and fairness. Courts must therefore adopt rigorous standards for evaluating the data sources and methodologies underlying AI systems before admitting such evidence.
These three characteristics- probabilistic reasoning, opacity, and data dependency collectively distinguish AI-generated evidence from traditional forms of evidence and complicate its admissibility under existing legal frameworks. Addressing these challenges requires not only doctrinal adaptation but also the development of new evidentiary standards that ensure transparency, reliability, and fairness in the use of AI within judicial proceedings.
Thus, AI-generated evidence represents a paradigm shift in evidentiary law, requiring a re-evaluation of foundational principles such as authenticity, reliability, and admissibility. It challenges the anthropocentric nature of traditional legal frameworks and necessitates the development of new doctrinal tools to address the complexities of machine-generated information.
LEGAL FRAMEWORK UNDER INDIAN LAW
The admissibility of AI-generated evidence in India is primarily governed by the provisions of the Indian Evidence Act, 1872 and its modern successor, the Bharatiya Sakshya Adhiniyam, 2023. While these statutes provide a framework for electronic and expert evidence, they were not designed to address the complexities introduced by artificial intelligence.
The admissibility of AI-generated evidence must be evaluated against the traditional requirements of relevance, authenticity, and reliability. These principles form the cornerstone of evidentiary law and are essential to ensuring fairness in judicial proceedings.
Relevance is the threshold requirement for admissibility. Under Sections 5 to 9 of the Indian Evidence Act, evidence must have a logical connection to the facts in issue. AI-generated evidence, such as predictive analytics or pattern recognition outputs, can be highly relevant in certain contexts, particularly in criminal investigations. However, relevance alone is insufficient; the evidence must also satisfy standards of authenticity and reliability.
Authenticity requires proof that the evidence is what it purports to be. In the context of AI-generated evidence, this involves verifying the integrity of the underlying data and the processes used to generate the output. The rise of deepfake technology has made this task increasingly difficult. Courts must ensure that AI-generated media has not been tampered with or fabricated.
Reliability is perhaps the most significant challenge in the context of AI-generated evidence. The Supreme Court in Ram Singh v. Col. Ram Singh laid down factors for evaluating the reliability of tape-recorded evidence, including accuracy, absence of tampering, and proper custody. These principles can be analogously applied to AI-generated evidence. However, the “black box” nature of AI systems complicates the assessment of reliability, as courts may not have access to the underlying algorithms or data.
Sections 65A and 65B of the Indian Evidence Act govern the admissibility of electronic records. Section 65B, in particular, mandates the production of a certificate to establish the authenticity of electronic evidence. The Supreme Court in Anvar P.V. v. P.K. Basheer clarified that compliance with Section 65B is a mandatory requirement for admissibility. This position was reaffirmed in Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, where the Court emphasized that electronic evidence without proper certification is inadmissible.
However, applying Section 65B to AI-generated evidence presents unique challenges. AI systems often involve complex data processing pipelines, making it difficult to identify a single “device” or “person” responsible for generating the certificate. Additionally, proprietary AI systems may not disclose their internal workings, thereby complicating the certification process.
Section 45 of the Evidence Act allows for expert opinions in matters requiring specialized knowledge. Courts have traditionally relied on human experts in fields such as forensics, handwriting analysis, and medical science. The question arises whether AI systems can be treated as “experts” under this provision. While AI tools may assist human experts, they cannot independently satisfy the requirements of expert testimony, particularly the ability to explain reasoning and withstand cross-examination.
The Bharatiya Sakshya Adhiniyam, 2023 attempts to modernize evidentiary law by expanding the scope of expert opinion. Section 39 of the Act includes “any field” of expertise, potentially encompassing AI systems. However, the Act does not explicitly address the admissibility of AI-generated outputs, leaving significant ambiguity.
In Tomaso Bruno v. State of Uttar Pradesh, the Supreme Court recognized the importance of electronic evidence in modern trials and emphasized the need for courts to adapt to technological advancements. This observation is particularly relevant in the context of AI-generated evidence.
In Sonu @ Amar v. State of Haryana, the Supreme Court held that objections to electronic evidence must be raised at the trial stage, failing which they may be deemed waived. This underscores the importance of procedural compliance in ensuring admissibility.
Thus, while existing statutory provisions provide a foundation for the admissibility of electronic evidence, they are insufficient to fully address the complexities introduced by AI. There is a pressing need for legislative clarification and judicial interpretation to bridge this gap, while traditional evidentiary principles provide a framework for assessing AI-generated evidence, their application requires careful adaptation to account for the unique characteristics of AI systems
JUDICIAL APPROACH IN INDIA
The Indian judiciary has played a pivotal and evolving role in shaping the admissibility framework for electronic evidence, which today serves as the closest doctrinal foundation for assessing the legal status of AI-generated evidence. Over the years, courts have actively interpreted statutory provisions, particularly those relating to electronic records, to ensure that evidentiary rules remain relevant in the face of rapid technological advancements. Through landmark judgments and progressive reasoning, the judiciary has attempted to strike a balance between procedural safeguards and the practical realities of digital evidence, thereby laying down guiding principles on authenticity, reliability, and admissibility.
Although courts in India have not yet directly adjudicated upon AI-generated evidence as a separate and distinct evidentiary category, their jurisprudence on electronic evidence offers valuable analogies and interpretative tools. The manner in which courts have dealt with issues such as digital authenticity, certification requirements, and the integrity of electronic records provides significant insight into how similar concerns surrounding AI-generated outputs may be approached in the future. In this sense, existing judicial precedents do not merely regulate electronic evidence but also indirectly shape the emerging legal discourse on artificial intelligence within the evidentiary framework.
The landmark judgment in Anvar P.V. v. P.K. Basheer fundamentally transformed the admissibility of electronic evidence by holding that compliance with Section 65B of the Indian Evidence Act is mandatory. The Court rejected the earlier position that electronic evidence could be admitted through secondary evidence provisions and emphasized the need for strict procedural compliance. This ruling established a rigid evidentiary threshold, prioritizing authenticity and reliability.
This position was reaffirmed and clarified in Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, where the Supreme Court held that failure to produce a Section 65B certificate renders electronic evidence inadmissible unless exceptional circumstances exist. The Court also clarified that such certificates can be produced at a later stage if justified, thereby introducing a degree of procedural flexibility.
In Tomaso Bruno v. State of Uttar Pradesh, the Supreme Court emphasized the importance of electronic evidence in modern criminal trials and observed that courts must adapt to technological advancements. This progressive approach indicates judicial openness to evolving evidentiary forms, including AI-generated evidence.
More recently, Indian courts have begun encountering issues related to AI-generated content with increasing frequency, reflecting the growing integration of artificial intelligence tools within legal and administrative processes. These developments mark an important shift, as courts are no longer dealing solely with traditional electronic evidence but are now confronted with more complex forms of machine-generated information. A notable and widely discussed instance occurred when the Gujarat High Court flagged the use of AI-generated fictitious case law in a quasi-judicial order, thereby bringing judicial attention to the potential dangers associated with uncritical reliance on such technologies.
The Court’s observation in this instance was particularly significant because it exposed how AI tools, if used without proper verification, can generate entirely fabricated legal citations that may appear credible on the surface but lack any concrete legal basis. This raises serious concerns about the reliability, authenticity, and integrity of AI-generated legal materials, especially when they are used in formal decision-making processes. The incident illustrates that AI systems, while powerful, are not infallible and may produce misleading or incorrect outputs if not carefully monitored.
Furthermore, this development underscores the broader risks associated with unverified AI outputs in the legal domain, including the possibility of judicial errors, erosion of trust in legal institutions, and compromise of procedural fairness. It also highlights the urgent need for heightened judicial vigilance, as well as the establishment of clear guidelines and safeguards to regulate the use of AI-generated content in legal proceedings.
Similarly, in Anil Kapoor v. Simply Life India, the Delhi High Court addressed the misuse of AI-generated deepfake content and recognized the threat posed by synthetic media to individual rights and legal processes. Although the case primarily concerned personality rights, it has broader implications for evidentiary law, particularly in relation to authenticity and manipulation.
The judiciary’s cautious approach reflects an underlying concern about the reliability and integrity of AI-generated evidence. Courts have consistently emphasized the need for procedural safeguards, particularly in relation to electronic evidence. However, the absence of specific guidelines for AI-generated evidence creates uncertainty and inconsistency in judicial decision-making.
Furthermore, the adversarial system relies heavily on cross-examination as a tool for testing the credibility of evidence. AI systems, however, cannot be cross-examined in the traditional sense, raising questions about how their outputs can be effectively challenged in court. This limitation necessitates the development of alternative mechanisms for evaluating AI-generated evidence.
In conclusion, while Indian courts have demonstrated adaptability in dealing with electronic evidence, the unique challenges posed by AI-generated evidence require a more nuanced and specialized approach. Judicial precedents provide a foundational framework, but legislative intervention and doctrinal innovation are essential to ensure consistency and fairness in the admissibility of AI-generated evidence.
CHALLENGES IN ADMISSIBILITY
The admissibility of AI-generated evidence presents a wide range of complex legal and technical challenges that cannot be adequately or effectively addressed within the existing evidentiary framework. As courts increasingly encounter technologically advanced forms of evidence, it becomes evident that traditional legal doctrines, which were designed for human-generated or static electronic records, are not fully equipped to handle the dynamic and evolving nature of AI-based outputs. The current framework lacks the necessary tools and standards to properly evaluate such evidence, thereby creating uncertainty in its treatment and admissibility within judicial proceedings.
These challenges arise primarily from the unique and distinctive characteristics of AI systems, which set them apart from conventional forms of evidence. One of the most significant issues is opacity, often described as the “black box” nature of AI, where the internal functioning and decision-making processes of the system are not transparent or easily explainable. In addition to this, AI systems are highly data-dependent, meaning that their outputs are directly influenced by the quality, scope, and potential biases present in the data used for training and operation. This raises serious concerns regarding accuracy, fairness, and reliability. Furthermore, AI-generated evidence is particularly susceptible to manipulation, whether through tampering with input data, altering algorithms, or generating synthetic content such as deepfakes. Collectively, these factors create substantial obstacles for courts in assessing the authenticity, reliability, and evidentiary value of AI-generated material.
One of the most significant challenges is the “black box” problem. AI systems, particularly those based on deep learning, operate through complex algorithms that are not easily interpretable. This lack of transparency makes it difficult for courts to assess the reliability of AI-generated outputs. In traditional expert evidence, the reasoning process can be scrutinized through cross-examination. However, AI systems do not provide a clear explanation of how conclusions are reached, thereby undermining the principle of procedural fairness.
Another major concern in the context of AI-generated evidence is the issue of bias and data quality, both of which significantly affect the reliability and fairness of such evidence. AI systems are typically trained on large volumes of data drawn from various sources, and these datasets may contain inherent biases arising from historical inequalities, incomplete representation, or flawed data collection methods. As a result, the outputs generated by AI systems are not neutral but may reflect and even amplify the biases embedded in the training data. This becomes particularly problematic in sensitive domains such as facial recognition technologies and predictive policing systems, where biased outcomes can have serious consequences for individuals and communities.
These biases can lead to discriminatory results, disproportionately affecting certain groups based on factors such as race, gender, or socio-economic background. In the legal context, such outcomes raise serious concerns about fairness and equal treatment before the law. The Supreme Court’s emphasis on fairness and equality in Maneka Gandhi v. Union of India highlights the constitutional importance of ensuring that all procedures affecting individual rights are just, fair, and reasonable. When AI-generated evidence is influenced by biased or unreliable data, it undermines these fundamental principles and calls into question the legitimacy of its use in judicial proceedings.
Furthermore, poor data quality- such as incomplete, outdated, or inaccurate datasets can further distort AI outputs, leading to erroneous conclusions. If courts rely on such flawed evidence, there is a risk of unjust outcomes, including wrongful accusations or decisions based on incorrect assumptions. Therefore, if AI-generated evidence is based on biased or low-quality data, its admissibility may not only be legally questionable but may also violate the principles of natural justice, which require fairness, impartiality, and the right to a fair hearing.
The proliferation of deepfake technology further complicates the issue of authenticity. AI-generated videos and audio recordings can be manipulated to create highly realistic but false representations. This poses a significant threat to the integrity of judicial proceedings. Courts must develop robust mechanisms to detect and verify such content.
Procedural challenges also arise in relation to Section 65B certification. As noted in Anvar P.V., electronic evidence must be accompanied by a certificate confirming its authenticity. However, in the context of AI systems, identifying the appropriate authority to issue such certification can be difficult, particularly when multiple entities are involved in data processing.
Additionally, there is a lack of standardized testing and validation procedures for AI systems. Unlike traditional forensic methods, which are subject to established protocols, AI systems often operate without uniform standards. This raises concerns about the consistency and reliability of their outputs.
The issue of accountability is also critical. In cases involving erroneous AI-generated evidence, it is unclear who should be held responsible is the developer, the user, or the system itself. This lack of accountability undermines the deterrent effect of legal sanctions.
In Selvi v. State of Karnataka, the Supreme Court emphasized the critical importance of safeguarding individual rights against the use of intrusive and unreliable scientific techniques in the process of investigation and evidence collection. The Court underscored that methods which interfere with personal liberty, mental privacy, or bodily integrity must be carefully scrutinized to ensure that they do not violate constitutional protections, particularly those guaranteed under Article 20(3) and Article 21. It highlighted that the use of such techniques, if not properly regulated, can compromise the voluntariness of evidence and undermine the fairness of the legal process.
This principle is particularly relevant and increasingly significant in the context of AI-generated evidence, which may involve extensive and sometimes invasive forms of data collection, processing, and analysis. AI systems often rely on large-scale personal data, including biometric, behavioural, or digital information, to generate outputs that may later be used in legal proceedings. Such practices raise serious concerns regarding privacy, consent, and the potential misuse of sensitive information. If these technologies operate without adequate safeguards, they may replicate the very risks identified by the Court in Selvi, thereby threatening individual autonomy and the integrity of the justice system.
In conclusion, the challenges associated with AI-generated evidence are multifaceted and require a comprehensive legal response. Without clear guidelines and safeguards, the use of such evidence may compromise the fairness and integrity of judicial proceedings.
COMPARATIVE ANALYSIS
A comparative analysis of international approaches to AI-generated evidence reveals a number of emerging trends and evolving legal principles that can significantly inform and guide the development of Indian evidentiary law. By examining how different jurisdictions are addressing the challenges posed by artificial intelligence in legal proceedings, it becomes possible to identify best practices, common concerns, and innovative regulatory responses. Such comparative insights are particularly valuable for a country like India, where the legal framework is still adapting to technological advancements and where there is a pressing need to modernize evidentiary rules in line with global developments.
While legal systems across jurisdictions vary in their regulatory structures, institutional capacities, and policy priorities, there is a clear and growing recognition worldwide of the need for specialized standards to deal with AI-generated evidence. Courts and lawmakers are increasingly acknowledging that traditional evidentiary rules are insufficient to address issues such as algorithmic opacity, probabilistic outputs, and data bias. As a result, many jurisdictions are moving towards developing tailored frameworks that emphasize reliability, transparency, accountability, and human oversight. These developments highlight a global shift towards more nuanced and technology-sensitive legal approaches, which can serve as an important reference point for strengthening and refining the Indian evidentiary regime.
In the United States, the admissibility of scientific and technical evidence is governed by the Daubert standard, established in Daubert v. Merrell Dow Pharmaceuticals, Inc. This standard requires courts to evaluate the reliability of expert evidence based on factors such as testability, peer review, error rates, and general acceptance within the scientific community. Although originally developed for human expert testimony, the Daubert framework has been increasingly applied to AI-generated evidence.
US courts have also begun to address issues related to algorithmic transparency. In State v. Loomis, the Wisconsin Supreme Court considered the use of a risk assessment algorithm in sentencing and emphasized the need for caution due to the lack of transparency and potential bias. This case highlights the importance of explainability in evaluating AI-generated evidence.
The European Union has adopted a more regulatory approach, focusing on transparency, accountability, and risk-based classification of AI systems. The proposed AI Act emphasizes the need for explainability and human oversight, particularly in high-risk applications such as law enforcement and judicial processes. This approach reflects a proactive effort to address the challenges posed by AI-generated evidence.
In contrast, India has not yet developed a comprehensive regulatory framework for AI. The existing legal framework relies on general principles of evidence and technology law, which may not be sufficient to address the complexities of AI systems. However, the judiciary’s emphasis on procedural safeguards and fairness provides a foundation for future development.
Comparative analysis suggests that India stands to benefit significantly from adopting and adapting elements of both the United States and European Union approaches to the regulation and evaluation of AI-generated evidence. Each of these jurisdictions offers distinct but complementary perspectives that can help shape a more comprehensive and effective evidentiary framework in the Indian context. In particular, the Daubert standard in the United States, with its strong emphasis on reliability, testability, peer review, and scientific validity, provides a useful model for assessing the credibility and evidentiary value of AI-generated outputs. Incorporating similar principles into Indian law could strengthen judicial scrutiny and ensure that such evidence meets rigorous standards before being admitted in court.
At the same time, the European Union’s approach, which places considerable emphasis on transparency, accountability, and human oversight, offers valuable guidance for regulatory reforms. The EU framework recognizes the risks associated with opaque and autonomous AI systems and seeks to mitigate them through clear obligations on developers and users. By integrating these principles, India can address concerns related to algorithmic bias, lack of explainability, and the need for responsible deployment of AI technologies.
Moreover, the global trend towards the development of specialized standards for AI-generated evidence highlights the urgency for India to establish its own coherent and context-specific legal framework. As jurisdictions worldwide move towards tailored regulatory mechanisms, it becomes increasingly important for India not to rely solely on traditional evidentiary rules that may no longer be adequate. By drawing on international best practices and adapting them to its unique legal and social context, India can ensure that its evidentiary law remains robust, forward-looking, and capable of responding effectively to rapid technological advancements.
AI-GENERATED EVIDENCE AS A DISTINCT EVIDENTIARY CATEGORY
IT INVOLVES AUTONOMOUS PROCESSING
The rapid integration of artificial intelligence into legal processes has created a pressing need to reassess traditional evidentiary classifications. AI-generated evidence, which includes outputs produced through machine learning algorithms, predictive analytics, and automated data processing systems, cannot be adequately understood or regulated within the confines of existing legal categories. While it is often tempting to treat such evidence as a subset of electronic records or expert opinion, this approach fails to capture its unique nature and operational complexity. A closer examination reveals that AI-generated evidence possesses characteristics that distinguish it fundamentally from both ordinary electronic evidence and traditional expert evidence, thereby necessitating recognition as a separate and distinct evidentiary category.
At the outset, it is important to understand why AI-generated evidence cannot be treated as ordinary electronic evidence. Traditional electronic evidence typically consists of data that is stored, transmitted, or reproduced in digital form, such as emails, documents, or video recordings. These forms of evidence are essentially passive in nature; they record or preserve information without altering its substantive content. The legal framework governing electronic evidence, particularly under provisions like Sections 65A and 65B of the Indian Evidence Act, 1872, is designed to ensure the authenticity and integrity of such records by focusing on issues like proper certification, chain of custody, and the reliability of the device used to produce the record.
IT PRODUCES PROBABILISTIC OUTPUTS
In contrast, AI-generated evidence is not merely a digital record but the result of autonomous processing carried out by complex algorithms. AI systems actively analyze data, identify patterns, and generate outputs that may not have existed in any prior form. This autonomous processing distinguishes AI-generated evidence from conventional electronic records, which do not involve independent decision-making or analytical functions. The system’s output is shaped by its internal architecture, training data, and programmed objectives, all of which operate without direct human intervention at the point of output generation. As a result, the evidentiary value of such outputs cannot be assessed solely on the basis of authenticity or certification, as is the case with ordinary electronic evidence.
IT ALSO DIFFERS FROM EXPERT EVIDENCE BECAUSE
Another critical distinction lies in the probabilistic nature of AI-generated outputs. Unlike traditional evidence, which is generally expected to establish facts with a reasonable degree of certainty, AI systems produce results based on statistical probabilities. For example, an AI-powered facial recognition tool may indicate a high likelihood that a particular individual matches an image, but it does not provide absolute certainty. This probabilistic approach is inherent to machine learning models, which rely on patterns and correlations rather than definitive conclusions. Consequently, treating such outputs as equivalent to conventional electronic evidence risks misinterpreting their evidentiary value and may lead to overreliance on results that are inherently uncertain.
In addition to differing from electronic evidence, AI-generated evidence also diverges significantly from traditional expert evidence. Under Section 45 of the Indian Evidence Act, expert evidence is based on the specialized knowledge and experience of human experts, who are capable of explaining their reasoning and methodology to the court. The credibility of expert testimony is tested through cross-examination, during which the expert’s qualifications, assumptions, and conclusions are scrutinized. This process ensures that the court can evaluate the reliability and relevance of the expert’s opinion in a transparent and accountable manner.
NO HUMAN REASONING
AI systems, however, do not possess human reasoning in the conventional sense. While they may simulate decision-making processes, their outputs are generated through algorithmic computations that are often not easily interpretable. The absence of human reasoning means that AI-generated evidence cannot be explained or justified in the same way as expert testimony. This limitation poses a significant challenge to the adversarial system, which relies on the ability to question and challenge evidence through cross-examination. Without a clear understanding of how an AI system arrived at a particular conclusion, it becomes difficult for courts to assess its reliability or for opposing parties to effectively contest it.
NO ACCOUNTABILITY
Furthermore, the issue of accountability presents a major obstacle in treating AI-generated evidence as expert evidence. Human experts can be held accountable for their opinions, whether through professional disciplinary mechanisms or legal liability for negligence or misconduct. In contrast, AI systems lack legal personality and cannot be held directly responsible for errors or inaccuracies in their outputs. While responsibility may theoretically be attributed to developers, operators, or users, the diffusion of responsibility across multiple stakeholders complicates the assignment of liability. This lack of accountability undermines the trustworthiness of AI-generated evidence and raises concerns about its use in judicial proceedings.
AI evidence represents a third category requiring: Given these distinctions, it becomes evident that AI-generated evidence does not fit neatly within existing evidentiary categories. Instead, it represents a third category that requires its own set of legal principles and standards. Recognizing AI-generated evidence as a distinct category would allow for the development of tailored rules that address its unique characteristics and challenges. Such an approach would also promote clarity and consistency in judicial decision-making, as courts would have a defined framework for evaluating this type of evidence.
VERIFICATION
One of the key requirements for this new category is verification. Unlike traditional evidence, where authenticity can often be established through straightforward means, AI-generated evidence requires a more comprehensive verification process. This includes assessing the integrity of the data used by the system, the reliability of the algorithm, and the conditions under which the output was generated. Verification mechanisms must be robust enough to detect errors, biases, and potential manipulation, ensuring that only reliable evidence is admitted in court.
HUMAN OVERSIGHT
Human oversight is another essential component in the evaluation of AI-generated evidence. While AI systems can process vast amounts of data with remarkable efficiency, they should not be allowed to operate without meaningful human supervision, particularly in legal contexts. Human oversight ensures that AI outputs are interpreted correctly and that any limitations or uncertainties are taken into account. It also provides a safeguard against blind reliance on technology, reinforcing the role of human judgment in the administration of justice.
TECHNICAL VALIDATION
Technical validation further complements verification and oversight by introducing standardized methods for assessing the performance and reliability of AI systems. This may involve independent testing, peer review, and adherence to established technical standards. Technical validation helps establish confidence in the system’s outputs and provides a basis for evaluating their evidentiary value. Without such validation, courts may struggle to determine whether an AI-generated output is sufficiently reliable to be admitted as evidence.
The unique features of AI-generated evidence- namely its autonomous processing, probabilistic nature, absence of human reasoning, and lack of clear accountability distinguish it from both electronic and expert evidence. These differences necessitate the recognition of AI-generated evidence as a separate evidentiary category, accompanied by specialized standards for verification, human oversight, and technical validation. By adopting such an approach, the legal system can better address the challenges posed by artificial intelligence while ensuring that the fundamental principles of fairness, reliability, and justice are upheld.
CRITICAL ANALYSIS
The emergence of AI-generated evidence exposes significant structural and doctrinal limitations within the Indian evidentiary framework. While existing provisions under the Indian Evidence Act, 1872 and the Bharatiya Sakshya Adhiniyam, 2023 provide a foundation for the admissibility of electronic and expert evidence, they are fundamentally inadequate to address the complexities introduced by artificial intelligence.
At the core of the issue lies the misfit between AI-generated evidence and existing legal categories. Traditionally, evidence is classified as oral, documentary, or expert. AI-generated outputs do not neatly fall into any of these categories. While they may be treated as electronic records under Sections 65A and 65B, this classification fails to capture the autonomous and analytical nature of AI systems. Unlike ordinary electronic records, which merely store or transmit information, AI systems actively process data and generate new outputs. This distinction is crucial, as it affects the evaluation of reliability and probative value.
Similarly, treating AI-generated evidence as expert opinion under Section 45 is doctrinally problematic. Expert evidence is premised on the assumption that the expert possesses specialized knowledge and can explain their reasoning to the court. In State of H.P. v. Jai Lal, the Supreme Court emphasized that expert opinion must be supported by clear reasoning and subject to judicial scrutiny. AI systems, however, lack the ability to articulate their reasoning in a manner comprehensible to courts. The “black box” nature of many AI models prevents meaningful cross-examination, thereby undermining the adversarial process.
Another critical issue is the probabilistic nature of AI-generated outputs. Unlike traditional evidence, which is expected to establish facts with a degree of certainty, AI outputs are inherently probabilistic and based on statistical correlations. This raises concerns about the standard of proof, particularly in criminal cases where the burden is “beyond reasonable doubt.” Reliance on probabilistic evidence may dilute this standard and increase the risk of wrongful convictions.
The constitutional implications of AI-generated evidence must also be considered. The Supreme Court in Maneka Gandhi v. Union of India established that any procedure affecting personal liberty must be “just, fair, and reasonable.” The use of opaque and potentially biased AI systems in judicial proceedings may violate this principle, particularly if individuals are unable to challenge the basis of the evidence against them. Similarly, the right to a fair trial under Article 21 may be compromised if courts rely on evidence that cannot be effectively scrutinized.
Bias in AI systems further exacerbates these concerns. AI models trained on biased datasets may produce discriminatory outcomes, particularly against marginalized communities. This raises issues under the constitutional guarantee of equality. The lack of transparency in AI systems makes it difficult to detect and address such biases, thereby undermining the legitimacy of judicial outcomes.
The issue of accountability also remains unresolved. In cases involving erroneous AI-generated evidence, it is unclear who should bear responsibility. Unlike human experts, who can be held accountable for negligence or misconduct, AI systems lack legal personality. This creates a gap in the enforcement of legal standards and weakens the deterrent effect of evidentiary rules.
Moreover, the procedural requirements under Section 65B are ill-suited to AI-generated evidence. The certification process assumes a linear chain of custody and control, which does not align with the distributed and dynamic nature of AI systems. This creates practical difficulties in ensuring compliance and may lead to the exclusion of potentially valuable evidence.
In light of these challenges, it is evident that the current evidentiary framework is insufficient to address the complexities of AI-generated evidence. A paradigm shift is required, moving beyond traditional categories and adopting a more nuanced approach that accounts for the unique characteristics of AI systems. This may involve the creation of a distinct evidentiary category for AI-generated outputs, accompanied by specialized standards for admissibility and evaluation.
RECOMMENDATIONS
Statutory Recognition of AI Evidence: A crucial step in addressing the challenges posed by artificial intelligence in legal proceedings is the statutory recognition of AI-generated evidence as a distinct evidentiary category. At present, Indian law primarily governs such evidence under the Indian Evidence Act, 1872 and the Bharatiya Sakshya Adhiniyam, 2023, where it is generally treated as a form of electronic record. However, this classification is inadequate because AI-generated evidence is not merely stored or transmitted data; it involves autonomous analysis and the creation of new outputs based on algorithmic processing. Therefore, explicit statutory recognition is necessary to ensure clarity, consistency, and proper evaluation in judicial proceedings.
The need for such recognition can be illustrated through practical examples. For instance, in criminal investigations, facial recognition systems may generate matches between suspects and surveillance footage. While such outputs can be highly useful, they are inherently probabilistic and dependent on training data. Treating them as ordinary electronic evidence under Section 65B may overlook issues such as algorithmic bias and error rates. Similarly, predictive policing tools that identify potential crime hotspots generate insights that are not direct evidence but algorithmic inferences, requiring careful legal scrutiny.
Judicial precedents on electronic evidence provide a foundation but also highlight the limitations of existing law. In Anvar P.V. v. P.K. Basheer, the Supreme Court emphasized the mandatory nature of certification for electronic records, ensuring authenticity and reliability. However, this framework does not address the complexities of AI systems, such as explainability and data bias. Likewise, in Arjun Panditrao Khotkar v. Kailash Kushanrao Gorantyal, the Court reaffirmed strict compliance with procedural requirements for electronic evidence. While these rulings strengthen evidentiary safeguards, they do not directly account for the autonomous and analytical nature of AI-generated outputs.
A more recent illustration of the risks associated with unregulated AI evidence is seen in the Gujarat High Court’s observation regarding the use of AI-generated fictitious case law in a quasi-judicial order. This incident underscores the dangers of relying on AI outputs without proper statutory guidance and verification mechanisms.
Therefore, statutory recognition of AI-generated evidence would enable the legislature to prescribe specific standards relating to transparency, reliability, and accountability. Such recognition would not only bridge existing legal gaps but also ensure that courts are better equipped to evaluate AI-driven evidence in a manner consistent with principles of fairness and justice.
Mandatory Explainability Standards: The introduction of mandatory explainability standards is essential for ensuring the fair and reliable use of AI-generated evidence in judicial proceedings. Explainability refers to the ability of an AI system to provide a clear, understandable account of how it arrives at a particular output or decision. In the legal context, this requirement is closely linked to principles of transparency, accountability, and procedural fairness, all of which are fundamental to the administration of justice.
AI systems, particularly those based on complex machine learning models, often function as “black boxes,” producing results without revealing the underlying reasoning process. This lack of transparency poses a serious challenge to courts, which are required to evaluate the credibility and probative value of evidence. Unlike human experts, who can be cross-examined and asked to justify their conclusions, AI systems do not inherently possess the ability to explain their reasoning in a comprehensible manner. As a result, relying on unexplained AI outputs may undermine the adversarial process and compromise the right to a fair trial.
The need for explainability can be linked to broader constitutional principles. The Supreme Court in Maneka Gandhi v. Union of India emphasized that any procedure affecting individual rights must be “just, fair, and reasonable.” In the context of AI-generated evidence, this implies that parties must have the opportunity to understand and challenge the basis of the evidence presented against them. Without explainability, such a challenge becomes practically impossible, thereby violating principles of natural justice.
A practical example can be seen in the use of AI-based facial recognition systems. If such a system identifies a suspect based on algorithmic matching, the court must be able to assess how the match was determined, including factors such as error rates, confidence levels, and the data used for training the model. Mandatory explainability standards would require developers and users of AI systems to disclose this information, enabling courts to make informed decisions regarding admissibility.
Furthermore, explainability promotes accountability by ensuring that those responsible for deploying AI systems can justify their outputs. This is particularly important in cases where erroneous AI-generated evidence may lead to wrongful decisions. By mandating transparency and disclosure, the legal system can mitigate risks associated with bias, errors, and misuse of AI technologies.
In conclusion, mandatory explainability standards are a critical safeguard for integrating AI into the evidentiary framework. They ensure that AI-generated evidence is not only technologically advanced but also legally reliable, transparent, and consistent with the principles of justice.
Certification Mechanism for AI Systems: A robust certification mechanism for AI systems is essential to ensure the admissibility, reliability, and integrity of AI-generated evidence in judicial proceedings. While existing law- particularly Section 65B of the Indian Evidence Act, 1872 and corresponding provisions under the Bharatiya Sakshya Adhiniyam, 2023, requires certification for electronic records, this framework is not adequately equipped to address the complexities of AI systems. Unlike ordinary electronic evidence, AI-generated outputs are the result of dynamic, multi-layered processes involving data collection, algorithmic training, and continuous updates. Therefore, a specialized certification mechanism tailored to AI systems is necessary.
Such a mechanism should go beyond merely verifying the authenticity of a device or record. It must encompass a comprehensive evaluation of the AI system itself, including its design, functionality, and performance. This would involve certifying the quality and integrity of the training data, the reliability of the algorithm, and the conditions under which the output was generated. For instance, in the case of a facial recognition system used as evidence, certification should confirm that the system has been tested for accuracy, that its error rates are within acceptable limits, and that it has not been trained on biased or incomplete datasets.
Judicial precedents on electronic evidence highlight the importance of certification but also reveal its limitations in the context of AI. In Anvar P.V. v. P.K. Basheer, the Supreme Court emphasized that certification is crucial for establishing the authenticity of electronic records. However, this approach assumes a relatively straightforward process of data generation and storage, which does not apply to AI systems. Similarly, Arjun Panditrao Khotkar v. Kailash Kushanrao reaffirmed the mandatory nature of certification, but did not address the challenges posed by autonomous systems.
A dedicated certification framework could involve the establishment of an independent regulatory or technical authority responsible for auditing and certifying AI systems used in legal contexts. This body would ensure compliance with standardized criteria relating to accuracy, transparency, and fairness. Additionally, certification should be periodic rather than one-time, given that AI systems evolve over time through updates and retraining.
In conclusion, a specialized certification mechanism for AI systems is indispensable for bridging the gap between traditional evidentiary rules and emerging technological realities. By ensuring that AI-generated evidence meets rigorous standards of reliability and accountability, such a mechanism would enhance judicial confidence and uphold the integrity of the legal process.
Judicial Training in AI: The effective integration of artificial intelligence into the evidentiary framework requires not only legal reform but also the development of institutional capacity, particularly within the judiciary. Judicial training in AI is essential to ensure that judges are equipped with the necessary knowledge and skills to understand, evaluate, and adjudicate cases involving AI-generated evidence. Without such training, there is a risk that courts may either over-rely on AI outputs without adequate scrutiny or reject them due to a lack of understanding, both of which can adversely affect the administration of justice.
AI systems operate on complex technical principles, including machine learning algorithms, data modeling, and statistical analysis. These concepts are not traditionally part of legal education or judicial training. As a result, judges may face difficulties in assessing critical aspects of AI-generated evidence, such as reliability, accuracy, bias, and error rates. For instance, when presented with evidence derived from facial recognition or predictive analytics, a judge must be able to evaluate not only its relevance but also the methodology and limitations of the system that produced it.
The need for judicial adaptability to technological advancements has been recognized by the Supreme Court in Tomaso Bruno v. State of Uttar Pradesh, where the Court emphasized the importance of keeping pace with modern methods of evidence collection and evaluation. This observation underscores the necessity of equipping judges with the tools to engage effectively with emerging technologies such as AI.
Judicial training programs should therefore include modules on the basic functioning of AI systems, their applications in legal contexts, and the potential risks associated with their use. Such programs may also involve collaboration with technical experts, workshops, and the development of practical guidelines for handling AI-generated evidence. Additionally, the appointment of court-appointed technical advisors or expert panels can assist judges in complex cases, ensuring informed decision-making.
Training in AI also promotes consistency in judicial outcomes. When judges possess a standardized understanding of AI technologies, it reduces the likelihood of divergent interpretations and enhances the predictability of legal decisions. Furthermore, it strengthens public confidence in the judicial process by demonstrating that courts are capable of addressing contemporary technological challenges.
In conclusion, judicial training in AI is a critical component of modernizing the legal system. By fostering technological literacy among judges, it ensures that AI-generated evidence is evaluated with the necessary rigor, thereby upholding principles of fairness, accuracy, and justice.
Independent Audit of Algorithms: The increasing reliance on AI generated evidence in legal proceedings necessitates the establishment of independent audit mechanisms to ensure the reliability, fairness, and integrity of algorithmic systems. An independent audit of algorithms refers to the systematic evaluation of AI models by external, unbiased experts to assess their performance, accuracy, and compliance with legal and ethical standards. Such audits are essential because AI systems, particularly those used in high-stakes contexts like criminal justice, can significantly influence judicial outcomes.
One of the primary concerns addressed by independent audits is the issue of bias and discrimination. As AI systems are trained on large datasets, they may inadvertently incorporate and amplify existing societal biases. Without proper oversight, this can lead to unfair or discriminatory results, especially in applications such as facial recognition or risk assessment tools. Independent audits can identify such biases by examining the training data, testing outputs across different demographic groups, and evaluating the system’s overall fairness.
Another important aspect of algorithmic auditing is the assessment of accuracy and error rates. Courts must be confident that AI-generated evidence meets acceptable standards of reliability before admitting it. Independent auditors can conduct rigorous testing to determine the system’s precision, false positive rates, and limitations under varying conditions. This is particularly relevant in light of judicial principles emphasizing the reliability of evidence, as seen in Ram Singh v. Col. Ram Singh, where the Supreme Court stressed the importance of accuracy and authenticity in evaluating recorded evidence.
Transparency is also enhanced through independent audits. Many AI systems are proprietary, and their internal workings are not publicly disclosed. Audits provide a mechanism for scrutinizing these systems without necessarily requiring full public disclosure of trade secrets. This balance between transparency and confidentiality is crucial for maintaining both innovation and accountability.
Furthermore, independent audits promote accountability by identifying the parties responsible for any deficiencies in the AI system. This is important in cases where erroneous outputs may lead to wrongful decisions. By establishing clear standards and regular audit requirements, the legal system can ensure that AI systems are continuously monitored and improved. independent audit of algorithms serves as a vital safeguard in the admissibility of AI-generated evidence. It enhances reliability, mitigates bias, and ensures that AI systems operate within acceptable legal and ethical boundaries, thereby strengthening the credibility of judicial processes
In conclusion, the challenges associated with AI-generated evidence require comprehensive reform of Indian evidentiary law. Although existing legal principles provide a foundation, they must be adapted to address the unique nature of artificial intelligence. This chapter proposes doctrinal, procedural, and institutional reforms to ensure fair admissibility.
First, statutory recognition of AI-generated evidence as a distinct category is necessary. Treating it as electronic or expert evidence is inadequate, as it fails to reflect its autonomous and analytical nature. A separate classification would allow tailored standards of transparency, reliability, and accountability.
Second, mandatory explainability standards must be introduced. Courts should require disclosure of training data, methodology, and limitations of AI systems to assess reliability. Transparency concerns have been highlighted in State v. Loomis, particularly regarding algorithmic opacity.
Third, a specialized certification mechanism is needed. Section 65B is insufficient for AI complexities. Certification should verify data integrity, algorithmic reliability, and compliance, ideally through an independent technical authority.
Fourth, judicial capacity-building is essential. Judges must be trained to understand AI evidence through programs, technical advisors, and guidelines. The Supreme Court in Tomaso Bruno v. State of Uttar Pradesh emphasized adapting to technological advancements.
Fifth, independent audits of AI systems should be mandated to assess accuracy, fairness, and compliance. Regular audits would reduce risks of bias and error, enhancing credibility.
Sixth, clear accountability rules must be established. Liability for erroneous AI evidence should be distributed among developers, deployers, and users to ensure enforcement and remedies.
Seventh, safeguards against deepfakes and synthetic media are necessary. Courts should require strict authentication, including forensic tools. In Anvar P.V., the Supreme Court stressed authenticity in electronic evidence, applicable to AI content.
Finally, India should adopt a comparative approach, drawing from the Daubert standard and EU transparency models to develop a balanced framework.
In conclusion, while AI-generated evidence poses challenges, it also offers reform opportunities. A structured approach will help integrate AI while preserving fairness and judicial integrity.
CONCLUSION
The admissibility of AI-generated evidence represents one of the most pressing challenges in contemporary Indian evidence law. While existing legal frameworks under the Indian Evidence Act, 1872 and the Bharatiya Sakshya Adhiniyam, 2023 provide a foundational structure for electronic and expert evidence, they remain insufficient to fully address the complexities introduced by artificial intelligence.
AI-generated evidence fundamentally differs from traditional forms of evidence due to its probabilistic nature, lack of transparency, and dependence on complex algorithms. These characteristics raise serious concerns regarding reliability, authenticity, and fairness—core principles that underpin the admissibility of evidence in any legal system. The absence of explainability in AI systems, commonly referred to as the “black box problem,” further complicates judicial evaluation, making it difficult for courts to assess the evidentiary value of such outputs.
Recent judicial developments in India demonstrate an increasing awareness of these challenges. Courts have shown caution in dealing with AI-generated materials, particularly where issues of authenticity and reliability arise. However, this cautious approach, while necessary, is not sufficient in the absence of a clear and coherent legal framework.
This paper has argued that AI-generated evidence should not be subsumed entirely under existing categories such as electronic or expert evidence. Instead, it should be recognized as a distinct evidentiary category requiring specialized standards. These standards must include requirements for transparency, reproducibility, validation, and human oversight.
Furthermore, legislative intervention is essential to establish clear guidelines governing the admissibility of AI-generated evidence. Such reforms should aim to balance the benefits of technological advancement with the need to preserve the integrity of judicial processes. Comparative analysis indicates that jurisdictions across the world are beginning to develop such frameworks, and India must not lag behind in this regard.
In conclusion, while AI holds immense potential to enhance the efficiency and accuracy of legal proceedings, its unregulated use as evidence poses significant risks to justice. The future of Indian evidence law will depend on its ability to adapt to these technological realities while upholding the fundamental principles of fairness, reliability, and due process.
BIBLIOGRAPHY
PRIMARY SOURCES
- Indian Evidence Act, 1872
- Bharatiya Sakshya Adhiniyam, 2023
- Information Technology Act, 2000
CASES
- Anvar P.V. v. P.K. Basheer, (2014) 10 SCC 473.
- Arjun Panditrao Khotkar v. Kailash Kushanrao, (2020) 7 SCC 1.
- Tomaso Bruno v. State of Uttar Pradesh, (2015) 7 SCC 178.
- Ram Singh v. Col. Ram Singh, 1985 Supp SCC 611.
- Sonu @ Amar v. State of Haryana, (2017) 8 SCC 570.
- Gujarat High Court observation on AI-generated case law, reported in Times of India, 2024.
- Anil Kapoor v. Simply Life India, CS(COMM) 652/2023 (Del HC).
- Maneka Gandhi v. Union of India, (1978) 1 SCC 248.
- Selvi v. State of Karnataka, (2010) 7 SCC 263.
- Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993).
- State v. Loomis, 881 N.W.2d 749 (Wis. 2016).
- State of H.P. v. Jai Lal, (1999) 7 SCC 280.
ARTICLES & PAPERS
- AI and admissibility challenges in Indian courts
- AI forensic evidence analysis
- Comparative AI evidence frameworks
- AI forensic admissibility under BSA
NEWS REPORTS
- Gujarat HC on AI-generated fake citations
- Deepfake concerns in Indian courts
- Emerging global AI evidence rules