From Code To Court - A Comprehensive Analysis Of AI Generated Evidence In The Criminal Justice System
Author: LawSphereX
Category: Criminal Law
Published:
Updated:
FROM CODE TO COURT: A COMPREHENSIVE
ANALYSIS
OF AI GENERATED EVIDENCE IN THE CRIMINAL JUSTICE SYSTEM
Avani Tungare, Symbiosis Law School, Pune
ABSTRACT
The law of evidence has undergone a radical change with the introduction of Artificial Intelligence (AI) into the criminal court system, presenting both benefits and challenges. AI has completely changed the way that evidence is gathered and examined. Examples of these technologies include automated document processing, facial recognition, and predictive policing. This study examines the definition of artificial intelligence (AI) and the standards used to identify evidence as AI-generated. It looks at how AI is now used in criminal justice, emphasizing how it affects accountability and the problems that come with it. The paper also discusses the objections raised against AI's application in law enforcement, in particular the worries over potential hazards and ingrained prejudices.
The paper also examines the requirements for AI-generated evidence's admissibility, evaluating how current legal systems either support or need these technological developments. The goal of the project is to improve the validity and equity of AI-based evidence in court processes by assessing the benefits and limitations of AI-driven evidence and suggesting innovative procedural protections. The article aims to further the current conversation on striking a balance between innovation and justice in a changing legal environment by providing this thorough analysis. Keywords: Evidence, Artificial Intelligence, admissibility, criminal justice Introduction ‘Change is the only constant’ The criminal justice system is no exception to being the witness of the transforming landscape vis-a vis the incorporation of Artificial Intelligence in Evidence Law.
AI systems have expedited the administration of criminal justice by assisting in predictive policing, facial recognition, and automated document processing, to name a few. 1 AI-driven technologies are already widely used by Law Enforcement Agencies (LEAs) for the collection and analysis of evidence. 2 AI-powered digital evidence has the power to completely change the vision of accountability. 3 However, to ensure its successful application, a number of challenges must be addressed. The availability of digital evidence can be impacted by elements including internet access, content removal policies implemented by platforms, and the difficulty of finding pertinent content amid the rapidly evolving social media landscape4. Furthermore, intentional misinformation efforts and concerns about AI-manipulated media are raising more and more doubts about the accuracy of digital information.
While AI is still being effectively implemented in many industries, there has been criticism around its use in the criminal justice system. Consequently, there are inadvertent biases built into the system, endangering a great number of people. In order to provide a thorough study of the advantages and disadvantages of this technological innovation, this research article will examine the evidential value of artificial intelligence (AI), how it is integrated into the criminal justice system, and the standards for admissibility of AI-generated evidence. Research Questions • What is the definition of artificial intelligence (AI) and how is AI evidence defined? 1 Ghanghash, Mr. S. (2022) ‘Artificial Intelligence in administration of justice in India: Challenges and opportunities’, Cyber Crime, Regulations and Security - Contemporary Issues and Challenges, pp. 178–187. doi:10. 55662/book.
2022ccrs. 008.
2 Id.
3 Nweke, O.C. and Nweke, G.I. (2024) Legal and ethical conundrums in the AI era
A multidisciplinary analysis, International Law Research. Available at: https://doi.org/10.5539/ilr.v13n1p1 (Accessed: 04 August 2024). 4 Peake, J. (2024) ‘Challenges of using digital evidence for war crimes prosecutions: Availability, reliability, admissibility’, AJIL Unbound, 118, pp. 57–61. doi:10.1017/aju.2024.5.
How is AI being integrated into the criminal justice system, and what are the potential threats and challenges associated with this integration? • What are the standards for the admissibility of evidence generated by AI in legal proceedings? Research Objectives • Explore and propose novel procedural safeguards for handling evidence generated by AI to ensure its reliability and fairness. • Critically evaluate existing standards and practices related to AI-generated evidence and refine them to address emerging challenges and ensure effective judicial processes. Legal Framework Definition and Scope of AI in the Legal System In general, Artificial Intelligence refers to the ability of a machine to practice deep learning and answer questions intelligently.
The High-Level Expert Group on Artificial Intelligence (AI HLEG), in its AI Ethics Guidelines, formally proposed to define AI as “systems designed by humans that, given a complex goal, act in the physical or digital world by perceiving their environment, interpreting the collected structured or unstructured data, reasoning on the knowledge derived from this data and deciding the best action(s) to take (according to predefined parameters) to achieve the given goal. AI systems can also be designed to learn to adapt their behaviour by analysing how the environment is affected by their previous actions.
As a scientific discipline, AI includes several approaches and techniques, such as machine learning (of which deep learning and reinforcement learning are specific examples), machine reasoning (which includes planning, scheduling, knowledge representation and reasoning, search, and optimization), and robotics (which includes control, perception, sensors and actuators, as well as the integration of all other techniques into cyber-physical systems). ”5 AI excels in pattern recognition within text files and documents, which is particularly useful in
5 (18 December 2018) A definition of ai. Available at:
https://ec. europa. eu/futurium/en/system/files/ged/ai_hleg_definition_of_ai_18_december_1. pdf (Accessed: 04 August 2024). organizing large volumes of data or handling complex cases. A notable application is ‘eDiscovery’, an automated process used in the United States to sift through electronic material for disclosure before legal proceedings begin. ‘eDiscovery’ leverages machine intelligence to identify key sections of vast datasets efficiently. The search parameters and algorithms are agreed upon by both parties and approved by a judge. This method of document analysis, recognized by courts in the United States and the United Kingdom, is faster and more accurate than manual file research. AI can offer valuable advice to individuals and potential litigants seeking resolution for their issues. It serves as an advisory tool for legal practitioners as well.
AI systems not only search for relevant information but also provide responses to specific queries. Users can then decide whether to follow these recommendations. This advisory capability helps individuals resolve many issues independently, potentially avoiding disputes and court proceedings. Additionally, AI can assist in formulating legal solutions, such as drafting petitions or summonses, thus aiding in standardizing judicial decisions. AI's ability to predict court rulings, often referred to as predictive justice, has garnered significant interest. However, this term is contentious, as the output of forecasting models is neither absolute nor entirely predictable. A more accurate term is "forecasting," which likens these predictions to weather reports rather than certainties.
Court proceedings, like weather patterns, can have unpredictable outcomes, with the uncertainty increasing as more information and variables are introduced. AI's appeal lies in its potential to reduce risk. In summary, artificial intelligence plays a dynamic and diverse role in the legal system, including prognostication, advising roles, and information organization. Even while AI may not be able to sit as a judge, it has already shown to be a useful tool for increasing the accuracy and efficiency of legal procedures. AI technology will probably become more and more integrated into the legal system as it develops, requiring constant modification and oversight to guarantee dependability and fairness.
AI and its Relationship with Evidence Law In India, the usage of AI has shown a significant correlation with the Bhartiya Sakshya Adhiniyam, 2023 (hereinafter, ‘BSA’), primarily through its impact on the admissibility and authentication of digital evidence. Several provisions within the BSA have gained prominence due to the evolving role of AI in collecting, preserving, and presenting evidence. One important provision is Section 636 of the BSA which deals with electronic evidence. As AI-driven technologies increasingly generate and store electronic data, this section has become pivotal in determining the admissibility of digital evidence in Indian courts. AI algorithms are being employed to retrieve and process electronic records, and Section 63 of the BSA sets guidelines for the certification and admissibility of such evidence.
AI's role in facial recognition and surveillance systems affects the interpretation of Section 39(2)7 of the BSA, which pertains to opinions of experts. Courts now need to consider the reliability and accuracy of AI-generated expert opinions when analyzing digital evidence, as AI systems can be considered expert witnesses in certain cases. The growth of AI has also necessitated amendments to the Indian Evidence Act to address emerging challenges and standards for AI-generated evidence. The synergy between AI technology and this legal framework illustrates the evolving relationship between technology and law in India.
Judicial Pronouncements and Comparative Study Indian and American Jurisprudence in context of Deciding the Admissibility Criteria for AI Generated Evidence Artificial intelligence (AI) and legal frameworks interact dynamically, which emphasizes the necessity of on-going modernization and adaptation to meet new opportunities and challenges in the gathering, preserving, and presenting of digital evidence. This requirement is reflected in the jurisprudence of both India and the United States, where instances highlight how the legal system must change to keep up with technological innovations. Indian Cases The judicial precedent set by Shafhi Mohammad and Ors. v. The State of Himachal Pradesh and Ors8. highlights the necessity of modernizing the investigative process.
While acknowledging that investigative agencies lacked the necessary equipment for videography, the court also emphasized the importance of its adoption, highlighting a larger point: in order to improve the quality and transparency of evidence, the legal system must keep up with
6 The Bharatiya Sakshya Adhiniyam, 2023, §63, No. 47, Acts of Parliament, 2023 (India).
7 The Bharatiya Sakshya Adhiniyam, 2023, §39(2), No. 47, Acts of Parliament, 2023 (India).
8 Shafhi Mohammad and Ors. v. The State of Himachal Pradesh and Ors., (2015) 7 SCC 178.
technological changes. The application of AI in evidentiary law is consistent with this principle. The court's acknowledgement emphasizes how crucial it is to be prepared and receive training before incorporating new technologies into an inquiry. In order to ensure adherence to legal standards and rules, law enforcement authorities and legal professionals need to obtain the requisite knowledge in using AI efficiently. Maintaining the integrity of the evidence and avoiding problems brought on by misuse or misinterpretation of AI are made possible by this training. The court also emphasized transparency and accountability when implementing new technologies. AI systems used in investigations must be transparent and capable of explaining their decision-making processes to ensure that generated evidence is fair, reliable, and admissible according to evidence law principles.
References to previous cases, such as Ram Singh and Ors v. Col. Ram Singh9 underline a broader principle in evidence law: the willingness to embrace new techniques and devices to improve evidence quality, provided their accuracy can be proven. This principle applies to both traditional evidence collection methods and emerging technologies like the Internet of Things (IoT). The overlapping of AI and evidence law in these references demonstrates the legal system’s adaptability to harness new technologies’ advantages while upholding accuracy and reliability standards. The court’s recognition in Shafhi Mohammad mirrors the willingness to adopt AI in evidence law, acknowledging AI's potential to enhance accuracy and efficiency, particularly in handling large data volumes, such as those generated by IoT devices. The case of State (N. C. T. of Delhi) vs.
Navjot Sandhu10 often referred to as the "Parliamentary attack case," offers insights into the admissibility of electronic records, specifically mobile data records. The court’s decision considered the admissibility of electronic evidence in the absence of strict compliance with Section 65B11 of the Indian Evidence Act, showcasing the legal
9 Ram Singh and Ors v. Col. Ram Singh, 1985 (Supp) SCC 611.
10 State (N. C. T. of Delhi) vs. Navjot Sandhu, 2005 11 SCC 600.
11 The Indian Evidence Act, 1872, §65B, No. 1, Acts of Parliament, 1872 (India).
system’s flexibility to incorporate modern technology while maintaining reliability and credibility standards in evidence presentation. The court’s approach in this case reflects the need for flexibility in evidence law concerning emerging technologies like AI. AI-generated evidence, often involving complex technology and data analysis, can be subjected to scrutiny and validation by competent witnesses knowledgeable about the AI systems used. The legal system’s adaptability is crucial to ensuring evidence law can accommodate AI while upholding established credibility, reliability, and fairness standards. A significant shift occurred with the Anvar P. V v. P. K Basheer12case, where the Supreme Court mandated strict compliance with Section 65B of the Indian Evidence Act for the admissibility of electronic evidence.
This ruling emphasized maintaining the integrity and authenticity of electronic records, crucial for next-generation evidence, particularly in cases involving computer and digital records. American Cases In the case of Williams v. City of Detroit13, a man was wrongfully identified through facial recognition and charged with a crime he did not commit. This incident highlights the potential biases in AI systems, particularly in facial recognition technology, which has incorrectly identified individuals of Asian or African American descent. The case emphasizes the need for rigorous vetting and validation of AI-generated evidence to prevent wrongful convictions. An analysis discussed in "Deep Fakes in Divorce Court: Manipulated Electronic Evidence and What to Do About It"14 highlights the fabrication of evidence through AI technology.
In one instance, a woman fabricated evidence showing domestic abuse during a custody battle, initially resulting in the court ruling in her favour. This underscores the necessity for thorough vetting of AI-generated evidence to ensure its authenticity and prevent manipulation. In Re: Celsius Network LLC15 the validity of AI-Privacy laws related to AI technology pose
12 Anvar P.V v. P.K Basheer, 2014 10 SCC 473.
13 Williams v. City of Detroit, No. 22-1344 (6th Cir. 2022)
14 Egan, M.J. (2020) FN 283 - deep fakes in Divorce Court
Manipulated electronic evidence and what to do about it: The recorder, Perma.cc. Available at: https://perma.cc/TWC6-8V9D (Accessed: 04 August 2024).
15 Re: Celsius Network LLC, 655 B.R. 301 (2023WL 8631746
challenges, particularly regarding the admissibility of AI-generated evidence. In the context of a murder investigation, questions arose about whether recordings from an Echo device could be used to identify the perpetrator. The case underscores the blurred lines between user privacy and the use of AI-generated evidence in legal proceedings. Produced evidence/data was questioned. The judge ruled the AI-produced data unreliable and inadmissible as part of the claim. This case illustrates that while AI has advanced significantly, it still produces mistakes, necessitating careful scrutiny of AI-generated evidence. An analysis in "Artificial Intelligence as Evidence" 16 by Paul W. Grimm, Maura R. Grossman, and Gordon V. Cormack discusses an instance where a voicemail transcript marked as spam led to misunderstandings.
This instance emphasizes how prone AI is to error and how crucial it is to guarantee the validity of evidence produced by AI. The plaintiff in Tice v. Amazon. com, Inc. 17 claimed that by recording private conversations, Alexa had violated the California Invasion of Privacy Act. The recordings were deemed admissible by the court since the plaintiff's spouse had consented to Alexa's terms and conditions. This example emphasizes how complicated privacy regulations are and how permission is needed for recordings made by AI. Customers filed a class action lawsuit against Amazon in Wilcosky v. Amazon. com, Inc. 18 alleging that the company had secretly recorded their discussions. The court's decision highlighted the complexity of privacy regulations pertaining to AI-generated evidence, as well as the necessity of unambiguous consent.
Therefore, the interaction between AI and evidence law in American and Indian jurisprudence illustrates how flexible the legal system is to capitalize on the benefits of emerging technology while maintaining standards of justice, precision, and dependability. Cases such as Williams v. City of Detroit and Tice v. Amazon. com, Inc. in the US and Shafhi Mohammad and Navjot Sandhu in India highlight the significance of openness, responsibility, and close examination when using AI into legal procedures. In order to meet the potential and problems posed by AIgenerated evidence, both jurisdictions emphasize the necessity of ongoing legal framework 16 Grimm, P. W. , Grossman, M. R. and Cormack, G. V. (2021) Artificial Intelligence as evidence. Available at: https://scholarlycommons. law. northwestern. edu/cgi/viewcontent. cgi? article=1349&context=njtip (Accessed: 04 August 2024).
17 Tice v. Amazon.com, Inc., 845 Fed. Appx. 535 (9th Cir. 2021).
18 Wilcosky v. Amazon.com, Inc., 517 F. Supp.3d 751 (N.D. Ill. 2021).
modernization and adaptation. This is necessary to guarantee that justice is still just, open, and efficient in the digital age. Critical Analysis Challenges of Relying on AI Generated Evidence and AI systems Determining the correctness and dependability of these AI-generated expert judgments is the main difficulty. Artificial intelligence (AI) systems lack human-like subjective experiences, prejudices, and the capacity to rationally justify their decisions. As a result, courts must take into account a number of factors: Training and Validation: The way AI systems are trained to produce expert opinions needs to be closely examined by courts. In order to do this, it must be determined if the AI was trained on a representative dataset and if the training data was gathered in accordance with moral and legal guidelines.
Making sure the models and techniques used for training are suitable for the particular domain in issue is crucial. Explicability and Transparency: The "black box" problem19—AI's lack of transparency—is one of the main issues with the technology. AI systems that are transparent and able to explain their decision-making processes must be mandated by courts. This openness makes it easier to comprehend how the AI came to its conclusions, which is crucial for stakeholders and legal experts to evaluate the accuracy and fairness of the opinions given by AI. To achieve these criteria, interpretable AI models and algorithms must be researched. Error Rates and Confidence Levels: Courts should inquire into the error rates and confidence levels associated with AI-generated opinions. This involves understanding the system's track record of accuracy and its capability 19 Quezada-Tavárez, K.
, Vogiatzoglou, P. and Royer, S. (2021) ‘Legal challenges in bringing AI evidence to the Criminal Courtroom’, New Journal of European Criminal Law, 12(4), pp. 531–551. doi:10. 1177/20322844211057019. to provide a measure of confidence in its conclusions. High error rates and low confidence levels can indicate situations where the AI may not be suitable for providing expert opinions or where additional corroboration is needed. Data Bias20 and Quality: Careful consideration needs to be given to the bias and quality of the data utilized to train AI systems. Expert views that are distorted or untrustworthy might result from biased or erroneous training data, and this can have serious legal ramifications.
Examining factors such data collection techniques, data representativeness, and the existence of any known biases or distortions in the training data are all part of the process of evaluating data quality. Courts should also be informed about on-going initiatives to reduce prejudice in AI systems and make sure that the AI being utilized conforms to industry best practices in this area. Corroboration: Although AI might offer insightful information, there might be circumstances in which human expert testimony or supporting data is required to support or refute conclusions drawn by AI. Conclusion Artificial intelligence (AI) in the criminal justice system offers enormous potential as well as difficult obstacles to overcome.
AI has the power to completely transform the way justice is administered through improved predictive policing, effective document processing, and advanced data analysis. But the adoption of it brings up serious questions regarding the validity, openness, and equity of evidence produced by AI. Artificial intelligence (AI) systems bring new difficulties that need to be addressed, but they may also process enormous volumes of data and offer insights that can help in legal procedures. The veracity of AI-generated evidence, for instance, may be impacted by problems with AI systems' validation and training procedures, decision-making processes' transparency, and potential biases in the data they employ. AI's place in evidence law will need to be carefully monitored as it develops to make sure that justice is enhanced rather than compromised.
20 Preeti Kushwah, Evaluating the Evidential Value of Evidence Generated by AI, 4 INDIAN J. L. & LEGAL RSCH. 1 (2022-2023).
The legal systems in India and the US are changing to take into account the necessity of keeping strict standards of proof while also keeping up with technological innovations. Legal precedents emphasize how crucial it is to incorporate AI into legal processes with accountability, accuracy, and transparency. Nonetheless, continuous modernization of legal rules is required to handle the particular difficulties presented by AI. Making sure AI technologies are used responsibly will be essential as they become more integrated into the legal system. This entails helping legal practitioners gain a deep grasp of AI in addition to improving legal norms. The criminal justice system can use AI to its advantage while upholding the values of justice and fairness by tackling these issues.
Suggestions By implementing the following suggestions, the legal system can better manage the complexities introduced by AI, ensuring that technological advancements contribute positively to the pursuit of justice. • Provide Complete AI Guidelines Clearly define the rules and requirements for using AI in legal situations. Determining the technical specifications for AI systems, such as transparency standards, training and validation processes, and accuracy thresholds, is part of this. These policies ought to be revised often to reflect changes in technology. • Boost Explainability and Transparency Encourage the creation and application of AI systems that can clearly explain the decisions they make.
Legal frameworks ought to mandate that artificial intelligence (AI) systems utilized for data gathering and analysis be able to generate comprehensible and verifiable justifications for their findings. • Put Strict Validation Procedures in Place AI-generated evidence should go through extensive validation before being admitted as evidence in court. This entails evaluating the correctness of the AI system, checking for biases in its training data, and making sure that any forecasts or assessments are supported by further data or professional testimony. • Handle Data Bias and Quality Create procedures to guarantee that the data representative, objective, and of excellent quality is used to train AI systems. To find and eliminate any potential biases, AI systems should be continuously audited and monitored.
• Promote Education and Knowledge Give legal professionals thorough instruction on the applications and constraints of artificial intelligence (AI) technology. How to assess AI-generated evidence, recognize any biases in it, and successfully incorporate it into legal procedures should all be covered in this course. • Encourage Judicial Oversight Courts ought to be actively involved in regulating the use of AI to the judicial system. It should be possible for judges and other legal professionals to assess AI-generated evidence objectively and make sure that its use complies with accepted legal norms. • Encourage Cross-Jurisdictional Collaboration To exchange best practices and create uniform approaches to AI in evidentiary law, encourage cooperation between the legal systems of various jurisdictions. This can assist in resolving issues and establishing uniform standards internationally.
Bibliography
Research Papers: Egan, M. J. (2020) FN 283 - deep fakes in Divorce Court: Manipulated electronic evidence and what to do about it: The recorder, Perma. cc. Available at: https://perma. cc/TWC6-8V9D (Accessed: 04 August 2024). Ghanghash, Mr. S. (2022) ‘Artificial Intelligence in administration of justice in India: Challenges and opportunities’, Cyber Crime, Regulations and Security - Contemporary Issues and Challenges, pp. 178–187. doi:10. 55662/book. 2022ccrs. 008. Gordon, J. -S. (2021) ‘Ai and law: Ethical, legal, and socio-political implications’, AI & SOCIETY [Preprint]. doi:10. 1007/s00146-021-01194-0. Grimm, P. W. , Grossman, M. R. and Cormack, G. V. (2021) Artificial Intelligence as evidence. Available at: https://scholarlycommons. law. northwestern. edu/cgi/viewcontent. cgi? article=1349&con text=njtip (Accessed: 04 August 2024). Kaur Sabherwal Kirandeep, J.
(2024) ‘The role of Artificial Intelligence in improving criminal justice system’, International Journal of Science and Research (IJSR), 13(4), pp. 999–
1002. doi:10.21275/sr24412153835.
Nweke, O.C. and Nweke, G.I. (2024) Legal and ethical conundrums in the AI era: A multidisciplinary analysis, International Law Research. Available at: https://doi.org/10.5539/ilr.v13n1p1 (Accessed: 04 August 2024). Peake, J. (2024) ‘Challenges of using digital evidence for war crimes prosecutions: Availability, reliability, admissibility’, AJIL Unbound, 118, pp. 57–61. doi:10.1017/aju.2024.5. Quezada-Tavárez, K., Vogiatzoglou, P. and Royer, S. (2021) ‘Legal challenges in bringing AI evidence to the Criminal Courtroom’, New Journal of European Criminal Law, 12(4), pp. 531–551. doi:10.1177/20322844211057019.
Statutes: Bhartiya Sakshya Adhiniyam, 2023 The Indian Evidence Act, 1872