AI, Indian Evidence Act, And Bhartiya Sakshya Adhiniyam - Navigating Legal Challenges In The Digital Age
Author: LawSphereX
Category: Technology, Cyber & Privacy
Published:
Updated:
AI, INDIAN EVIDENCE ACT, AND BHARTIYA SAKSHYA
ADHINIYAM: NAVIGATING LEGAL CHALLENGES IN THE
DIGITAL AGE
Abinaya. S, Tamil Nadu National Law University, Tiruchirappalli
ABSTRACT
This research delves into the dynamic intersection of Artificial Intelligence (AI), the Indian Evidence Act, and constitutional rights in the context of legal proceedings in India. It scrutinizes the admissibility and reliability of AIgenerated evidence, such as forensic analysis and predictive algorithms, within Indian Courts, exploring their alignment with established principles of admissibility and relevancy. Furthermore, it examines the implications of AI in evidence collection, focusing on concerns related to privacy, selfincrimination, and protection against self-testimony guaranteed under Articles 20(3) and 21 of the Indian Constitution. The study also recognizes the need for legislative reforms to accommodate the integration of AI and the Internet of Things (IoT) into the Indian legal framework.
These reforms aim to establish specific guidelines and safeguards, ensuring the admissibility, reliability, and safeguarding of constitutional rights in an increasingly digitized legal landscape. Keywords: Artificial Intelligence (AI), Evidence, Constitutional Rights Admissibility of Evidence, Legislative Reform
I. INTRODUCTION
Artificial Intelligence (AI) has emerged as a transformative force in almost every facet of modern society, revolutionizing industries, services, and operations. The legal field is no exception, with AI technologies being increasingly integrated into various aspects of legal practice, including research, document review, and translation. AI has made significant inroads into the legal field, particularly through predictive algorithms and forensic analysis. Predictive algorithms use AI and machine learning to assist legal professionals in tasks like legal research, document review, predicting case outcomes, and risk assessment These technologies provide data-driven insights and streamline decision-making.
In the realm of forensic analysis, AI plays a pivotal role in digital forensics, fingerprint and facial recognition, voice and audio analysis, pattern recognition, and ballistics analysis. These tools help law enforcement and legal experts in investigations and evidence collection. The Indian Evidence Act, a cornerstone of the Indian legal system, provides a robust framework for the admission and exclusion of evidence in court proceedings. As AI applications become more prevalent in evidence collection and presentation, the compatibility of these technologies with the principles outlined in the Indian Evidence Act is called into question. The intersection of AI and the Indian legal system raises intricate concerns related to constitutionality, legal admissibility, and the protection of individual rights.
This article delves into the conundrums that emerge as AI becomes entwined with the Indian Evidence Act, exploring how it may impact constitutional safeguards and existing legal norms. Central to this examination is the critical analysis of Article 20 (3) of the Indian Constitution and other pertinent laws, as well as their implications for AI-driven evidence in Indian courts. By addressing these issues, we aim to shed light on the evolving landscape where technology meets the law, offering a comprehensive perspective on the intricate relationship between AI and the Indian legal system.
RESEARCH GAP
Artificial intelligence has become the new normal. It is being inculcated in each and every sector and the legal field is of no exclusion. Courts have started using AI in translation and other documentary works. The issue arises as to how the integration of artificial intelligence
into legal processes, including evidence gathering and presentation, impacts the constitutional and legal rights of individuals in India, particularly concerning Article 20 (3) of the Indian Constitution and existing legal frameworks.
RESEARCH OBJECTIVES
1. To analyze the conundrum of the usage of artificial intelligence and its applicability in
the legal field.
2. To assess the admissibility and reliability of AI-generated evidence in Indian Courts
within the framework of the Indian Evidence Act.
3. To examine the legal implications of using AI in evidence collection with a focus on
privacy, self-incrimination, and constitutional rights.
4. To explore the necessity and suggest amending Indian legal frameworks to provide
guidelines and safeguards for the use of AI and IoT in legal proceedings, addressing admissibility, reliability, and constitutional considerations.
RESEARCH QUESTIONS
1. Whether AI-generated evidence, such as forensic analysis and predictive algorithms,
can be considered admissible and reliable in Indian Courts, and how does this align with the principles of admissibility and relevancy as established in the Indian Evidence Act?
2. Whether the use of AI in evidence collection and presentation raises concerns about
privacy, self-incrimination, and protection against self-testimony, as guaranteed under Articles 20 (3) and 21 of the Indian Constitution, and how are these challenges addressed within the legal framework?
3. Whether there is a need for comprehensive amendments to Indian Legal frameworks,
including the Indian Evidence Act and Criminal Law, to establish specific guidelines and safeguards for the use of AI and IoT in legal proceedings, addressing issues related to admissibility, reliability, and constitutional rights?
METHODOLOGY
The research methodology employed in this study is doctrinal in nature. Both Primary and secondary sources are used to collect the data, such as statutes, books, journals, articles, web sources, etc. Both critical and exploratory thinking will be involved.
II. AN INITIAL OVERVIEW OF ARTIFICIAL INTELLIGENCE
Artificial Intelligence (hereinafter AI) refers to the hypothetical ability of a computer to equal or surpass human capabilities in activities that involve cognitive functions, such as perception, comprehension, generation of language, logical thinking, creativity, and emotional understanding. 1 It signifies the ambitious objective of pushing the boundaries of computation, a broad concept encompassing software and engineering2 and can be summarised as “whatever tasks computers cannot perform... until they are able to”. AI depending on its ability and relationship with human performance is divided into two kinds. , i. e. , Narrow or weak AI and General or Strong AI. The former uses purpose-built hardware and/or software that emulates or works for the betterment of human performance at a specific, well-defined task.
3 The latter denotes the potential of a computer with or surpasses human proficiency across wide range of cognitive functions, encompassing tasks such as self-sustainability, in addition to other cognitive abilities. 4 Foundational AI technologies address abstract problems like classification (categorization), ranking (ordering by value), and regression (quantitative estimation). They can determine if something belongs to a category, rank items by importance or estimate specific values. It helps in summarizing data and predicting future outcomes. To achieve this, a supervised machine learning algorithm is trained to make distinctions between the objects or concepts, it consecutively builds a model which classifies the examples fed into the system.
Most of the models despite providing a categorical answer, perform regressions or estimation, probability, or confidence scores to provide a conclusion. The two vital supervised machine learning approaches are active learning and reinforcement learning, where the algorithm autonomously
1 A.M. Turing, I- Computer Machinery and Intelligence, 59 MIND 433, 460 (1950).
2 Engineering, Merriam-webster.com dictionary, [https://perma.cc/V3FR-Q4ZT].
3 Jake Frankenfield, Weak AI: Examples and Limitations, Investopedia, (Jan 1, 2022).
4 Strong AI, IBM Clous education, Perma | What is Strong AI? | IBM.
selects training examples, striving to strike a balance between exploration (maximizing learning) and exploitation (problem-solving). 5 Deep Learning refers to a fusion of multiple learning algorithms that jointly handle feature analysis and complex tasks like classification, ranking or regression. These are realized as multi-layered neural networks. Each layer takes raw data and processes it passing the result to the next layer. Every layer examines distinct features and adapts its model based on training data, enhancing overall efficiency. It has yielded significant advancements in speech and image recognition along with fact-based question answering. In the current arena, the majority of AI systems are built on machine learning (hereinafter MI) algorithms. It doesn’t follow pre-programmed data rather it allows systems to learn from data, experiences, and examples.
The permutation of Machine learning is as follows:
1. Supervised learning: Using labelled data as training material, the system learns from
the structure of the data and applies it to forecast new data categories
2. Unsupervised learning: It looks for similarities or shared characteristics between data
points.
3. Reinforcement learning: This method centers on monitoring and learning about the
norms and consequences of the surroundings by interacting with the environment and acquiring experience.6 Predictive AI Models offer foresight into forthcoming occurrences, biometrics aid in identification, and AI transcription services transform audio into written court evidence. Judges encounter difficulties when determining the acceptability of such evidence, as it involves issues related to reliability, transparency, interpretability, and bias in such evidence. This becomes even more challenging with the usage of generative AI systems causing misinformation and disinformation.7 5 Thomas Osugi et al., balancing exploration and exploitation: A new algorithm for active machine learning.
6 Daniel Seng, Artificial Intelligence and evidence, (2021) 33 SAcLJ 241.
7 ‘How to determine the admissibility of AI-generated evidence in courts?’, UNESCO, 26 July, 2023.
III. ADMISSIBILITY AND RELIABILITY OF AI AND AI-GENERATED
EVIDENCE IN INDIAN COURTS
3. 1- Admissibility of Scientific and Electronic Evidence: The following will explain the position and treatment of electronic and scientific evidence according Indian Evidence Act, of 1872. 3. 1. 1—Electronic Evidence The 21st century witnessed a global technological revolution, with computers accessible to individuals, transforming human actions through Information Technology. This led to the growth of the Cyber world and the need for updated laws in India concerning electronic evidence. The IT Act, 2000, and amendments based on UNCITRAL’s Model Law enable the recognition of digital evidence, complemented by changes in the Indian Evidence Act, IPC, and other laws forming the legal framework for electronic transactions. 8 Evidence that is stored or transferred in an electronic binary format that has probative value is referred to as electronic evidence.
9 It include evidence from digital devices other than computers, such as cell phones, GPS, and digital audio and video, as well as evidence from digital devices used in telecommunications and electronic multimedia. Evidence refers to documentary evidence which includes electronic records produced for the inspection of the Court. 10 In order to regulate the admissibility of electronic evidence, Sections 65A and 65B were added, and the phrase "Content of documents or electronic records were substituted" was added. The Evidence Act now has new provisions, Section 22-A, that address the relevance of oral testimony with reference to electronic record contents. It says that unless the veracity of the produced electronic records is in doubt, oral admissions concerning their contents are irrelevant.
11 Cyber forensics now plays a crucial role in the process of presenting digital evidence in a way that is compliant with the law in any court or administrative proceeding. 8 Vivek Dubey, Admissibility of electronic Evidence: An Indian Perspective, Forensic Research & Criminology International Journal, eISSN:2469-2794.
9 Information Technology Act, Sec 79 A, 2000.
10 Indian Evidence Act, Sec, 1872, amended by virtue of Sec 92 of IT Act, 2000.
11 Indian Evidence Act, Sec 22-A, 1872.
The exclusive means to establish the authenticity of electronic records or evidence is by presenting the original electronic storage device as primary evidence in Court, or a copy of it by way of secondary evidence u/s 65A and 65 B of the Indian Evidence Act. Items classified as secondary evidence must be accompanied by a section 65 B certificate that was obtained at the time the document was taken; if this requirement is not met, the secondary evidence relative to the electronic record is inadmissible. 12 The Courts have held that in the present era, nothing can be worse than denying discoveries and not making use of recent technological and other developments to prove the guilt of the accused. Just relying on conventional methods discourages the usage of efficient scientific aid. 13 3. 1.
2- Scientific Evidence Scientific evidence is regarded as one of the most important and necessary components to guarantee the administration of justice. It offers realistic remedies for the current criminal situation. There is a greater push for the usage of forensic techniques for criminal cases in India. 14 Scientific evidence, generally, comprises findings from scientific investigations employed to support or refute theories and hypotheses. In the realm of criminal cases, scientific evidence serves the purpose of aiding judges in comprehending and establishing the facts. 15 This type of evidence compasses various elements such as Polygraph tests, brain mapping, Lie detection16 DNA evidence, finger and footprints17, voice identification, bullet striations, gunshot residue, hair and skin analysis, voice recognition, tire imprints, and autopsy reports.
All these forms of evidence are relevant to the decision-maker. In the legal context, it is categorized as opinion evidence by an expert. 18 The evidence has to be authentic, reliable, and believable in order to be admissible in a court of law. Courts give due importance to the opinions of experts. Nevertheless, expert testimony is not considered substantial proof and is 12 Sri. P. Rajendra Prasad, Recent trends in recording and admissibility of evidence, district. ecourts. gov. in
13 Som Prakash v. State of Delhi AIR 1974 SC 989
14 Government of India, Ministry of Home Affairs, Press Information Bureau, Forensic Science Capabilities; Strengthening the time-bound and Scientific Investigation. . 15 Dr. Kusum Chauhan, ‘Admissibility and evidentiary value of scientific Evidence: Legislative and Judicial Approach in India”, International Journal for Research Trends and Innovation, Vol 8, Issue 1,2023.
16 Selvi v. State of Karnataka 2010 7 SCC 263.
17 Pritam Singh v. State of Punjab; Mohd. Aman v State of Rajasthan AIR 1997 SC 2960.
18 Indian Evidence Act, Sec 45, 1872.
typically used to support or refute the credibility of oral testimony.19.It basically serves as advising material20 for the court, which evaluates the opinion impartially and uses the criteria to help judges form their own decisions.21.
3.2 Admissibility and Reliability of Artificial Intelligence as Evidence in the
Court of Law The primary form of AI-generated evidence is electronic in nature. As discussed in 3. 1. 1, it is inherently susceptible to intentional or unintentional alterations or deletions. Consequently, it is crucial to ensure the trustworthiness and reliability of both the evidence itself and the system that store, process, or analyses it. 22 While there may be a transition from automated systems to AI systems within investigative and enforcement agencies for tasks like policing23, intelligence and information gathering24, evidential analysis,25 and employment by the courts for administration, and resolution of disputes26, the lawyers and courts will continue to place paramount importance on ensuring the trustworthiness and reliability of both the evidence and the underlying system.
It is important to distinguish between using information, including data generated by AI systems, to establish facts in legal proceedings and using such information to manage and oversee various legal aspects of litigation, investigations, and legal processes.
3.2.1 Admissibility of Artificial Intelligence
As discussed in the Introductory chapter, the fundamental of AI technology is the data and algorithm or algorithms that were created or developed for its governance and performance. Algorithms mean a set of rules or procedures for solving a problem or accomplishing an end. In order to determine the admissibility of AI Technology it is pertinent to define the problem for which the AI was designed to solve. This will make the Court clear in terms of its validity, 19 State of U.P. v. Krishna Gopal AIR 1988 SC 2154; Siddheswar Prasad Singh v. Baba Sundari Pramanick AIR
1978 Cal 4; Ram Chandra v. State of U.P AIR 1957 SC 381.
20 Malay Kumar Gnaguly v. Dr. Sukumar Mukherjee 2009 9 SCC 221.
21 Ramesh Chandra Agarwal v. Regency Hospitals Ltd 2009 9 SCC 709.
22 Stephen Mason & Daniel Seng eds, ‘Electronic evidence’, Institute of Advanced Legal Studies for the SAS Humanities Digital Library, School of Advanced Study, University of London, 4th edn, 2017.
23 Bridges, R (On Application of ) v. The Chief Constable of South Wales Police (2020) IWLR 672.
24 Patrik Perrot, ‘What about AI in criminal intelligence? From Predictive Policing to AI Perspectives’ 2017 16 European Police Science and Research Bulletin 65.
25 Own Bowcott & Hannah Devlin’ Police trial AI Software to help process mobile phone evidence, The
Guardian (27 May 2018). 26 Adam Harkens, ‘Fairness in Algorithmic Decision-Making; Trade-offs, policy choices, and procedural Protections’ 2019.
reliability and substantial functional creep if any. To ensure that the evidence is admissible, the proponent of the evidence must be aware of the design's goal and purpose. Relevance is not a concept that is abstract; rather, it is only relevant inasmuch as it can support or refute facts that are essential to the case's outcome. “The Daubert Factors could also be employed for admission of the trustworthiness of AI, it is of following:
1. Whether there is a known error rate
2. Whether the methods used are generally acceptable as reliable within the
relevant scientific or technical community that is familiar with the methodology
3. Whether the methodology has been subject to peer review by others
knowledgeable in the specified field.
4. If the standard procedures or protocols are applicable to the methodology,
whether they are compiled with.”27 This does not require perfection and the question that must be decided in each and every case is whether the evidence is sufficiently valid and reliable for the purpose it is being offered.
3.2.2 Presumption of Reliability:
In general. If electronic evidence is pertinent to the circumstances at hand and meets the requirements outlined in Sections 65 A and 65 B of the Indian Evidence Act, it may be admitted in a court of law. Common Law assumes that computer systems are trustworthy, and if there is no evidence to the contrary, the court will assume that mechanical instruments were functioning properly at the relevant time. 28 The issue of who shoulders the burden of demonstrating or refuting the validity of evidence produced or assisted by AI systems emerges. This is relevant because, even to their programmers and operators, AI systems, and some ML algorithms in particular, would function in ways that are opaque and not immediately apparent.
It must be handled differently than other electronic evidence because many judges view relevant evidence regardless of expert advice regarding the credibility of computer-electronic evidence. Countries
27 Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579, 593-94 (1993).
28 ‘Evidence in Criminal Proceedings: Hearsay and related topics, The Law Commission, United Kingdom (CP No 138, 1997) para 13.13.
such as Singapore29, and other common law countries have elevated the presumption into legal presumption that reallocates the burden of proof, not on the proponent but on its opponent. The relevant scientific evidence must demonstrate that the equipment or method in question is one of the kinds that, when utilised correctly, often makes or accurately communicates an electronic record in order to satisfy the Evidential Presumption. The prosecution must provide evidence, not just assertions, to back up its claims that the system it has chosen is "robust" or "high quality”. 30 The authenticity and trustworthiness of the system can ex-facie come from users who have used the system. 31 The proponent still bears the legal burden of demonstrating the machine's reliability as well as the authenticity, integrity, and reliability of the evidence if the presumption is strongly established.
This is because there will be an evidential presumption that the system in question is reliable. 32 As stated earlier the complexity of the AI system makes this procedure even more complicated because the traditional methods of testing software for mistakes such as unit tests and unit test platforms are unlikely to be very helpful, especially pertaining to corner cases or exceptional circumstances wherein the errors embedded in the AI system will activate only on exceptional circumstances. For instance, In the case of Uber's autonomous vehicle that killed a pedestrian, the main problem was the AI System for environmental perception which had difficulty in classifying the victim as a pedestrian, rather considering it to be an unknown object, than as a vehicle, etc due to the defect on the collision detection logic and other human interventions.
Henceforth, the proof of the reliability of AI Systems will therefore involve a combination of many mechanisms.
3.2.3 Categories and Treatment of AI Evidence:
human input and records it for subsequent retrieval. The Generally inputs retrieved are used as testimony. considered a Hearsay Category 2 Evidence is a product of automation and is not used Outside the testimonially. Considered as real evidence or evidence purview of the produced mechanically without any human Hearsay Rule. intervention33. Example ANPR (Automatic number plate recognition) evidence. This does not guarantee that it is reliable or accurate. Challenges can be made in the form of authentication. Category 3 Evidence produced comprises of mix of human-supplied Input and data-processed output which operates without Hearsay or not human intervention. As ML is trained based on humandepends on a labelled data it can operate autonomously therefore it factual basis. will invariably fall under this category. Expert opinion is allowed.
system. Conclusions are reached when data is input, with intricate "black box" computations occurring in between. This 'black box' is frequently difficult for the courtroom to understand in addition to being inaccessible to experts. This opacity, combined with its mechanical look and seemingly simple output, can give rise to the mistaken belief that AI is essentially dependable, objective, and assured. It is crucial to open this "black box" in order to guarantee a just and knowledgeable verdict for the defendant. Furthermore, as AI outputs are frequently unexplainable, the question of how defendants might defend themselves and refute AI-generated evidence becomes crucial. 3.2.4.2 Discrimination and Bias - Moreover, algorithmic decisions can stem from
incomplete and, therefore, unreliable data, which might be compromised by cyberattacks, biased, or simply erroneous. A hasty deployment of evolving technology without due consideration would result in problematic outcomes and public resistance to its adoption by the legal system, as the risk of malfunctions remains ever-present. Consequently, one of the most formidable challenges in effectively integrating AI into the criminal justice system is the mitigation of biases to which AI is susceptible. These biases, if left unaddressed, can contribute to flawed and unjust judicial decisions when taken into account by the judge or jury. In reality, these processes harbor concealed subjectivities and errors that often go unnoticed and uncorrected, potentially enabling the concealment of unlawful discrimination behind layers of complexity. 34 3. 2. 4.
3 Lack of Accountability- Moreover, when data is initially collected or
generated, it's common to encounter human errors in the collection or interpretation process.35 These errors can manifest during the data's training phase or even during the program's ongoing development. However, establishing accountability becomes challenging, as identifying the individual responsible for errors in machine learning systems, where the decisionmaking process is often inscrutable to computer scientists, is a daunting task. Additionally, AI experts and companies often argue in favor of AI systems,
34 A Roth, ‘Trial by Machine’ (2016) 104(5), Georgetown Law Journal 1245-306, 1269-70.
35 WA Logan and AG Ferguson, ‘Policing Criminal Justice Data’, Minnesota Law Review 549-615, 2016.
particularly machine learning ones, by highlighting their capacity to evolve in unforeseeable ways, driven by their autonomous and self-learning nature. Consequently, they contend that no programmer can be held responsible for the system's evolution.
IV. CONSTITUTIONAL RIGHTS IN THE AGE OF AI: BALANCING
TECHNOLOGY AND LEGAL PROTECTIONS
AI’s impact on fundamental rights can be assessed from multiple perspectives. One approach involves examining the use of AI and AI-generated evidence as primary sources of evidence. Another perspective delves into the role of AI in gathering and processing additional evidence.
4.1 The development of AI in assisting the electronic and scientific evidence
Forensic DNA testing has significantly impacted the criminal justice system in recent decades, enabling the detection and processing of biological materials like blood, saliva, semen, and skin cells left at crime scenes. Advances in DNA technology have increased sensitivity, allowing the examination of even low-level or degraded DNA evidence that was previously unusable. This progress has led to the re-examination of old DNA evidence from unsolved cases, such as sexual assaults and cold homicide cases. However, this enhanced sensitivity has brought new challenges. Highly sensitive methods can detect DNA from multiple contributors or individuals unrelated to the crime, creating complexities in DNA mixture interpretation. This necessitates the separation and identification of individual profiles to generate valuable leads for law enforcement.
Artificial intelligence (AI) holds promise in addressing this challenge. DNA analysis produces complex electronic data with patterns that may be beyond human analysis capabilities but prove valuable as AI systems become more sensitive. Researchers at Syracuse University, in collaboration with forensic science organizations, explored a machine learning-based method for mixture deconvolution. By combining human analyst approaches with data mining and AI algorithms, they sought to separate and identify individual DNA profiles, minimizing the limitations of using a single method in isolation. While ongoing research is essential and many factors can influence the accuracy of individual DNA donor identification, AI technology
shows the potential to aid these intricate analyses.36 Secondly, AI algorithms are being developed to discover the pattern of signatures in the gunshot analysis, differentiate muzzle blasts from shock waves, and determine shot-to-shot timings which could help law enforcement in the investigation.37
4.2 AI and ML Devices: Unravelling the Conundrum of Primary Evidence
The advent of AI technology has significantly enhanced the ease and convenience in people's lives. Contemporary devices like smart speakers, and home automation systems such as Alexa, Google Assistant, and Siri, not only respond to vocal commands but also have the capability to record them. These devices function as invaluable assistants and companions to individuals in the modern age. Consider a scenario where a mini-player is present at the time of a criminal offense, such as a homicide. It may have captured the conversation between the victim and the perpetrator, rendering it potentially admissible as evidence in a court of law. 38 To be deemed admissible, it must meet the criteria of relevance to the matter in question and satisfy the evidentiary elements as per Bentham's concept of "factum probandum" and "factum probans.
" Following due procedures for admission, it could be legally admissible. Alternatively, if the Mini player recorded the dying declaration of the victim, Section 32 (1)39 would mandate its legal admissibility for further investigation. The issue of whether illegally obtained evidence from the Mini player can be admitted arises, and it invokes legal precedents like R v. Maqsud Ali40 and Buxton v. Cumminq41. It raises the question of whether the doctrine of "fruits of the poisonous tree" applies, rendering such evidence inadmissible. Furthermore, concerns arise regarding potential violations of an 36 “A Hybrid Machine learning approach for DNA Mixture Interpretation”, Syracuse University, National Institute of Justice. 37 Christopher Rigano, ‘Using Artificial Intelligence to address criminal justice needs, National Institute of Justice, R.
‘Development of Computational Methods for the audio analysis of Gunshots’, Cadre Research Labs, LLC, NIJ.
38 Saloni Jaiman & Anushman Mishra, ‘Whether AI Mini-Players could be considered valid evidence in the eyes of Law’ The Society for Advancement of Criminal Justice, 2022.
39 Indian Evidence Act, Sec 32 (1), 1872.
40 R v. Maqsud Ali 2 All E.R. 464 (1965) para 16, p 63.
41 Buxton v. Cumminq 71 S.J. I 232, para 45, p 67.
individual's privacy, as protected under Article 2142, and the right against self-incrimination, governed by Article 20 (3)43 of the Indian Constitution. This question also arises with respect to other AI and ML devices as mentioned in Chapter 2 of the research paper. Henceforth, it must be solved on a case-to-case basis depending upon the category of AI or ML Evidence.
4.3 Whether AI-elicited evidence infringes upon the Right to Privacy enshrined
in Article 21 and contravenes the protections against self-incrimination as outlined in Article 20 (3) of the Indian Constitution? Article 21 of the Indian Constitution states that “No person shall be deprived of his life or personal liberty except according to a procedure established by law”. It has been held by the Court of Law that the Right to privacy is recognized under Article 21 and it is not absolute in nature and must comply with the public interest. 44 “According to Article 21, an invasion of privacy must be justified on the basis of a law which stipulates a procedure which is fair, just, and reasonable. The law must also be valid with reference to the encroachment on life and personal liberty under Article 21.
45 Also “when conflict arises as to the right to privacy and the right to a fair trial46 (all stages- stage of the investigation, inquiry, trial, appeal, revision, and retrial) both under the ambit of Art 21, the right to privacy may have to yield to the right to a fair trial”47 Merely because evidence is collected illegally doesn’t render it inadmissible in the Courts. 48 Though the Right to Privacy is a serious concern in the digital arena, the mere usage of AI-based evidence will not amount to a breach of Privacy. Article 20(3) upholds the Right against self-incrimination i. e. , “No person accused of any offense shall be compelled to be a witness against himself”. In the Selvi v.
State of Karnataka case, the Supreme Court held that “various evidence-gathering techniques such as Narcoanalysis, brain mapping, lie detection, and polygraphy tests are violative of Article 20 (3) and Article 21 of the Constitution. These tests cannot be regarded as completely truthful and
42 Indian Const, art 21, 1950.
43 Indian Const, art 20 cl 3, 1950.
44 Ritesh Sinha v. State of Uttar Pradesh 2 SCC 357 (2013)
45 Justice K.S. Puttaswamy v. Union of India 10 SCC 1 (2017), para 47 p.95.
46 Seesa Hemachandra Sashissal v. State of Maharashtra AIR 1246 (SC 2001), para 35 p.54; Dagdu v. State of Maharashtra CrLJ 1206 (1977); A.R. Antulay v. Avdesh Kumar AIR 1701 SC 1992, para 35 p.64.
47 Deepti Kapur v. Kunal Jhulka SCC On Line 672 (Del, 2020), para 38 p.76.
48 Pooran Mal v. The Director of Inspection (Investigation), New Delhi 1SCC 345 (1974).
reliable, hence there are chances of false results and a scope of errors. It comes in conflict with the standard of proof ‘beyond reasonable proof’ which is an essential element of criminal trials. However, on voluntary administration of the test by the accused, the results are however could be admitted in consonance with Section 27 of the Indian Evidence Act”.49 With the support of AI devices and their methodologies further substantiating the reliability, authenticity, and trustworthiness of the techniques wouldn’t amount to self-incrimination, but rather make sure the trial is speedy and proved beyond a reasonable doubt. As discussed in 4.1 it would make the scientific evidence stronger and reliable making significant changes in the criminal trial.
V. SUGGESTIONS AND CONCLUSION
Artificial Intelligence has become an integral part of our daily lives, and its growing presence is undeniable. While the existing Indian Evidence Act covers electronic evidence, it confines itself to binary formats, leaving a significant gap in addressing AI-related evidence. In contrast, the recently introduced Bhartiya Sakshya Adhiniyam in 2023 offers a more comprehensive perspective by encompassing information stored in semiconductor memory and various communication devices. 50 However, the Bhartiya Sakshya Adhiniyam does not explicitly address the use of Artificial Intelligence and Machine Learning. As we've discussed earlier, AI devices like Mini players, which record both written and spoken communication, fall within the scope of this legislation.
It is worth noting that one of the inherent challenges of AI is its susceptibility to discrimination and bias, primarily stemming from the data it is trained on. Consequently, this poses a potential risk of discrimination within AI-driven evidence. To ensure that AI's admissibility and reliability are effectively addressed in this new digital era, it is crucial for the legislature to proactively consider its implications. Neglecting this aspect would grant the judiciary extensive discretionary powers and hinder the legal field's adaptation to evolving technology. It is incumbent upon policymakers to acknowledge and accommodate these technological advancements, thereby enhancing the legal framework's responsiveness to the challenges and opportunities presented by Artificial Intelligence.
49 Selvi v. State of Karnataka (2010) 7 SCC 263.
50 The Bharatiya Sakshya Bill, Ministry of Home Affairs, https://prsindia.org/billtrack/the-bharatiya-sakshyabill-2023.
The research asserts that the admissibility of AI and AI-generated evidence in a court of law is contingent upon meeting the stipulated criteria outlined in Section 65A and 65B, Section 45, and other relevant provisions of the law. To secure admissibility, it is incumbent upon the proponent to establish the reliability and trustworthiness of such evidence. The utilization of AI in the process of evidence collection and presentation should not be seen as a blanket violation of the Right to Privacy. The Right to Privacy, while fundamental, is not an absolute right and must be balanced with other legal considerations. In cases where AI or ML models compel the accused to self-incriminate, it could potentially infringe upon Article 20(3) of the Indian Constitution.
Furthermore, it is imperative for the legislature to take a proactive stance in addressing the intricacies of Artificial Intelligence and Machine Learning within the legal framework. This regulatory intervention is necessary to prevent these matters from being solely at the discretion of the judiciary. Many Western and Common Law countries have already embraced the integration of AI into their criminal proceedings and trials. Therefore, it is crucial for Indian courts to seize this opportunity and establish a legal framework that accommodates the evolving role of AI in the legal domain.
BIBLIOGRAPHY
STATUTE
1. The Indian Evidence Act, 1871
2. The Indian Constitution, 1950
3. The Data Protection Act, 2023
4. Bhartiya Sakshaya Adhiniyam, 2023
COMMITTEE REPORTS
1. Report of Committee-D On cyber security, safety, legal and ethical issues, Artificial
Intelligence Committees Reports, Ministry of Electronics and Information Technology, Government of India.
2. AI technology and the justice system: Lords Committee report, UK Parliament, 2022
CASE LAWS:
1. Sanjaysinh Ramrao Chavan vs. Dattatray Gulabrao Phalke MANU/SC/0040/2015.
2. Anvar P.V. v. P.K. Basheer and Ors 2014 10SCC 473;
3. State v. Mohd.Afzal and Ors 2003 DLT 385
4. Jagdeo Singh v. The State and Ors. MANU/DE/0376/2015
5. Navjot Sandhu v State (2015) 3 SCC 123.
6. Som Prakash v. State of Delhi AIR 1974 SC 989
7. Selvi v. State of Karnataka 2010 7 SCC 263.
8. Pritam Singh v. State of Punjab
9. Mohd. Aman v State of Rajasthan AIR 1997 SC 2960
10. State of U.P. v. Krishna Gopal AIR 1988 SC 2154
11. Siddheswar Prasad Singh v. Baba Sundari Pramanick AIR 1978 Cal 4
12. Ram Chandra v. State of U.P AIR 1957 SC 381.
13. Malay Kumar Gnaguly v. Dr. Sukumar Mukherjee 2009 9 SCC 221.
14. Ramesh Chandra Agarwal v. Regency Hospitals Ltd 2009 9 SCC 709.
15. Sapparo Maru v. State of Liberty (1968) I WLR 739.
16. R v. Maqsud Ali 2 All E.R. 464 (1965) para 16, p 63.
17. Buxton v. Cumminq 71 S.J. I 232.
18. Ritesh Sinha v. State of Uttar Pradesh 2 SCC 357 (2013)
19. Justice K.S. Puttaswamy v. Union of India 10 SCC 1 (2017)
20. Seesa Hemachandra Sashissal v. State of Maharashtra AIR 1246 (SC 2001)
21. Dagdu v. State of Maharashtra CrLJ 1206 (1977)
22. A.R. Antulay v. Avdesh Kumar AIR 1701 SC 1992
23. Deepti Kapur v. Kunal Jhulka SCC On Line 672 (Del, 2020)
24. Pooran Mal v. The Director of Inspection (Investigation), New Delhi 1SCC 345
(1974).
ARTICLES AND WEB SOURCES
1. Paul W. Grimm, Maura R. Grossman & Ors, ‘Artificial Intelligence as Evidence’
Northwestern Journal of Technology and Intellectual Property, Vol 19, Issue 1, Article 2, 2021.
2. Fakultas Hukum, ‘Application of Artificial Intelligence as Evidence in Indonesian
Courts’, International Journal of Social Science and Business, Volume 5, Number 2, 2021, pp.180-184.
3. Steven J. Murdoch and Ors, ‘The Sources and characteristics of electronic evidence
and artificial intelligence’, University of London Press (2021), Jstor.
4. How to determine the admissibility of AI-generated evidence in Courts, United
nation educational, Scientific and Cultural Organisation, https://www.unesco.org/en/articles/how-determine-admissibility-ai-generatedevidence-courts.
5. Amen Jauhar, ‘Facial Recognition in law enforcement is the litmus test for India’s
commitment to Responsible AI for all,’ https://www.orfonline.org/expertspeak/facial-recognition-in-law-enforcement-is-the-litmus-test-for-indias/.
6. Saloni Jaiman and Anshuman Mishra, ‘ Whether AI Mini-Players could be
considered valid evidence in the eyes of law’, the Society for Advancement of Criminal Justice, 2022. https://www.nujssacj.com/post/whether-ai-mini-playerscould-be-considered-valid-evidence-in-the-eyes-of-law
7. Roshan Jain, ‘Admissibility of Electronic Record in India’, Manupatra, 2021.
8. Information and Communication Technology, Pathshala, MHRD Govt of India,
MODE OF PROOF OF ELECTRONIC EVIDENCE (inflibnet.ac.in).
9. P. Rajendra Prasad, Recent trends in recording and admissibility of evidence,
WORKSHOP IV NOTES (2).pdf (ecourts.gov.in)
10. Dr. Kusum Chauhan, ‘Admissibility and Evidentiary Value of Scientific Evidence:
Legislative and Judicial Approach in India’, IJRTI, Vol 8, Issue 1, 2023.
11. Shreya Jain & Ujaala Jain, Expert Evidence, Bharati Law Review, April-June 2018.
12. Chaudhary & Arvindeka, ‘Admissibility of Scientific Evidence under Indian
Evidence Act 1872’, shodhganga, 2014.
13. Christopher Rigano, “Using Artificial Intelligence to Address Criminal Justice
Needs,”
NIJ
Journal 280, January 2019, https:// www.nij.gov/journals/280/Pages/using-artificial intelligence-to-address-criminaljustice-needs.aspx
14. Daniel Seng, Stephen Mason, ‘Artificial Intelligence and Evidence’, 2021 33 SAcLJ
241. , Vol 21:3, p 919- 958.
15. Eftychia Bampasika, ‘Artificial Intelligence as Evidence in Criminal Trial’,
ethics7.pdf (ceur-ws.org).
16. Patrick W. Nutter, ‘Machine Learning Evidence: Admissibility and Weight’, Journal
of Constitutional Law