Integration of Artificial Intelligence in Forensic Evidence: A Critical Appraisal of Admissibility Reliability and Due Process in Bangladesh

Abstract

Applications of Artificial Intelligence to forensic evidence can potentially revolutionize Bangladesh’s criminal justice system with its capability to solve cases beyond the current norms. But this piece critically analyzes the vast gap between Bangladesh’s outdated legal framework, primarily the Evidence Act of 1872, and the colossal level of AI technology. The current framework fails to provide modern standards of admissibility and credibility of algorithmic evidence, particularly where concerns like the black box problem, algorithmic bias, and court literacy are concerned. The research highlights that the non-transparency and non-contestability of AI-based evidence are threats to fundamental due process rights as well as the risk of wrongful convictions. More far-reaching legislative and judicial reform is needed to create a flexible framework that will ensure that technological advancement ensures constitutional fairness and a right to a fair trial.

Introduction

Forensic science has always been the clincher in criminal investigation, offering crime-solving essentials and conviction acquisition tools. Every year, forensic methods such as ballistics matching, DNA typing, and fingerprint recognition continue to develop in leaps and bounds. Forensic science, however, is on the verge of transformation with the looming mainstreaming of Artificial Intelligence (AI) technologies. With the unique potential to process gigantic volumes of information, identify patterns, and draw conclusions that would potentially escape human examiners’ notice, the intersection of artificial intelligence and forensic science has the potential to transform crime scene examination.^[1]

Forensic science is assuming more significance in Bangladesh, which is a nation challenged by a complex criminal justice system plagued with sluggish judicial processes, congested courts, and mishandling of evidence. Forensic science methods, including fingerprinting, computer forensics, and Deoxyribonucleic Acid (DNA) analysis, have emerged as the very tools required as crime becomes increasingly advanced, providing fair, transparent, and unbiased results in criminal investigations.^[2]

With such forensic science capability and the latest artificial intelligence (AI) technology to uncover the truth of criminal investigations, this article uncovers the grand disconnect between such technology and Bangladesh’s obsolete legal system. The argument is that the current law, founded largely on the Evidence Act 1872, is not capable of dealing with the new legal and ethical issues arising from AI. It does not establish clear, modern standards for algorithmic evidence admissibility, ensure reliability against algorithmic prejudice and cover-ups, and protect due process rights. This calls for a wholesale overhaul of the legislative and judicial machinery to make sure that technological developments serve constitutional justice rather than erode it.

Research Methodology

Research Design

A variety of data gathered from various legislation, court rulings, and internet sources, including books, government websites, research articles, newspapers, and reports, are used in this study, which employs a qualitative research methodology and analytical technique. This approach ensures a thorough understanding of the subject and the incorporation of other points of view.

Research Objectives

The primary objective of this study is to determine the effectiveness of our current laws in addressing the challenges posed by the increasing use of artificial intelligence (AI) techniques in conjunction with conventional forensic evidence in criminal cases. To enable a more thorough analysis, two additional objectives are as follows: RO1: to evaluate the current legal framework governing the admissibility and reliability of AI-based forensic evidence; RO2: to highlight issues with due process and the potential for an incorrect conviction due to algorithmic bias and opacity.

Research Question

How effective is the current legal system of Bangladesh in addressing the core challenges of admissibility and reliability relating to the integration of AI tools with forensic evidence in criminal adjudication?

Conceptual Framework

This study is built on the concept of the intersection of forensic evidence and AI in the criminal justice system of Bangladesh via the prisms of admissibility and reliability. Any questions regarding the admissibility and reliability of any forms of evidence are solely governed by the Evidence Act of 1872. Forensic evidence,^[3] which includes blood, semen, DNA, finger impressions, etc, has become an important tool in proving or disproving any fact in criminal cases. Artificial intelligence (AI) in forensics refers to the use of AI technology, such as computer vision and machine learning, to enhance forensic investigations in many ways, including facial recognition, video or picture breakdown, and predictive evidence flagging.^[4]

Furthermore, the admissibility of AI evidence has to go through various standards (Daubert^[5] & Frye^[6]) of relevance, authenticity, and reliability; none of these standards are expressly stated in the Evidence Act of 1872. Besides, due process under Articles 27, 31, and 35 of the Constitution of Bangladesh demands that technological tools operate within guarantees of a fair trial. The structure thus places the research in the broader debate on technological innovation against constitutional safeguards and judicial integrity in contemporary criminal justice.

Legal Frameworks on Forensic Evidence and AI in Bangladesh

The following provisions of the given legislation are pertinent to the use of forensic evidence in Bangladesh, particularly in criminal cases:

(i) The Evidence Act 1872

The admission of forensic evidence used to be exclusively governed by Section 45. Expert opinions on issues about forensic evidence, probable science, art, identification, finger impressions, handwriting, etc., fall under this category. This breadth may be expanded. Section 73B was added to the Act with the 2022 amendment, enabling the court to compare forensic or physical evidence with other types of evidence.

(ii) The Code of Criminal Procedure 1898

Using forensic evidence is covered in Sections 174, 176, and 510. The applicability of the inquest inquiry and medical report is covered in Section 174. The magistrate is expressly authorised by Section 176 to order exhumation to determine the cause of death. It has to do with forensic medicine. Last but not least, section 510 gives the court the authority to take chemical examiner and serologist results into account. The reports of these experts may be used as evidence in court by court.

(iii) The DNA Act 2014

This Act brought about a change in Bangladesh’s criminal justice system. This Act requires judges and magistrates to have the authority to order a DNA test if it is relevant to the case and the results are acceptable as forensic evidence in court. This law specifies when and how DNA testing should be used, and it depends on the court to justify the situation.

The use of forensic evidence is explained in depth in this important legislation. The use of forensic evidence is evident in several more statutes in addition to these, which are given in the following table.

Table: I Laws on Forensic Evidence

Source: Musferat & Ummay^[7]

However, artificial intelligence evidence and its admissibility in court have not been addressed by any of these statutes. However, section 65B of the Evidence Act provides for the inclusion of AI evidence in the category of digital recordings.

The National Artificial Intelligence Policy of 2024 was drafted by Bangladesh in 2024. This proposed legislation aims to ensure “the ethical application of AI as we move towards achieving a Smart Bangladesh by 2041” and is aimed to promote the nation’s economic growth, social progress, and national security.^[8] Particularly when compared to more extensive and model regulatory frameworks such as China’s AI law and the European Union’s (EU) Artificial Intelligence Act, several aspects of the proposed regulation are deficient.

Judicial Rulings

In Bangladesh, there is currently no special legal or judicial treatment for evidence produced or processed by AI. While several court decisions have addressed the reliability and admissibility of digital and forensic evidence, none have addressed the incorporation of artificial intelligence into evidence.

Nonetheless, a few other nations have rendered court rulings in this regard, like in the case of Kohls v. Ellison^[9] federal district court in Minnesota rejected an expert opinion that contained artificial intelligence-generated citations because it compromised the expert’s credibility. The court emphasised the need to verify AI-generated data and the need for experts and attorneys to use their own discretion..

In Daubert v. Merrell Dow Pharmaceuticals,^[10] for the trial court to establish whether scientific data is admissible, the Supreme Court established requirements. They are whether or not the theory or method has been tested, whether or not it has been published and peer reviewed, whether it has a known error rate, whether there are standards governing its use that exist or which have been maintained, and whether or not it has gained general acceptance in the pertinent scientific community.

Benefits of AI for Forensic Interpretation

AI can greatly improve forensic evidence interpretation in a number of ways, resolving many of the issues and difficulties that come with conventional forensic techniques. AI systems can analyse evidence using consistent, standardised processes. They don’t experience the same problems that human specialists do, such as prejudice, stress, or exhaustion. This can guarantee that the evidence is interpreted uniformly in various cases. AI is faster and more efficient than humans in analysing and identifying intricate patterns in forensic data. This can be especially helpful for jobs that require comparing big datasets, including fingerprint and DNA analysis. AI can assist in lowering the possibility of human mistakes in the gathering and examination of forensic evidence by automating repetitive and regular processes. This might raise the general standard and dependability of the evidence.^[11]

The capacity of AI to automatically indicate potentially pertinent data is one of its most important contributions to forensics. Thousands of files, emails, call logs, documents, and media can be analysed by AI algorithms to find patterns or correlations that need further research. AI-powered facial recognition systems can identify people in public and private databases by analysing video from CCTVs, drones, or cellphones. AI can analyse video footage frame by frame and identify faces, objects, gestures, and questionable activity. Physical gear is only one aspect of contemporary forensic investigations. AI is being used more and more to analyse data that has been taken from social media platforms, cloud storage accounts, messaging applications, and mobile devices and SIM cards.^[12]

Admissibility and Reliability Challenges of AI in Forensic Analysis

The application of artificial intelligence (AI) in forensic evidence marks a revolutionary time of breakthroughs and unprecedented challenges. If AI-generated forensic evidence proliferates and becomes more like non-AI evidence, courts will have to address thorny problems regarding its validity, reliability, and admissibility. The following discussion will highlight the challenges concerning the admissibility and reliability of AI in forensic analysis:-

(i) Lack of Authentication

There are several challenges in the chain of custody and authenticity of AI-generated evidence. The opaque nature of AI processes makes it difficult for parties to demonstrate the integrity of evidence produced by a computer rather than a human, even if the legal system requires strict criteria for evidence admission. In the high-stakes world of forensic science, poor data quality can seriously undermine the reliability of artificial intelligence’s results.^[13]

(ii) Issues relating to Accuracy & Reliability

The accuracy and reliability of integrating AI into forensic evidence are one of the main issues. AI-generated information may not have a clear provenance or may be the product of intricate procedures that are challenging to audit, in contrast to traditional forensic evidence, which is frequently directly traceable and validated by human sources. This begs the question of how to demonstrate that such proof hasn’t been altered or produced using skewed data.

(iii) The Black Box Challenge

Because AI applications at times act as “black boxes,” it may be difficult to explain how they produce certain outcomes. These technologies infer patterns and correlations in data, which at times may produce erroneous conclusions, unlike conventional scientific approaches that seek causality. Biases may also result from incomplete or skewed datasets, casting doubt on the validity of these instruments. Vendors also frequently hide source code from the courts, claiming it is a trade secret or proprietary. For example, these reasons were used to deny access to TrueAllele’s source code in the 2021 New Jersey State vs. Pickett case. After it was discovered that the Office of the Chief Medical Examiner was employing unvalidated software for DNA analysis, the City of New York pursued legislation regarding automated decision systems utilised by its agencies.^[14]

(iv) Algorithmic Bias

One of the biggest obstacles to integrating artificial intelligence technology into forensic applications has been identified as algorithmic bias. Bias may be introduced in a number of ways and can have a significant effect on the results of studies. It is crucial to make sure that the datasets used to train the model are sufficiently diverse to reflect the whole population that the forensic science services are made up of.^[15]

(v) AI as an Expert Opinion

There are a number of difficulties with using AI as an expert opinion, including algorithmic bias from faulty data, a lack of explainability and transparency when decision-making is opaque, problems with data security and privacy, and a lack of responsibility when mistakes are made.

Given the complexity of many AI systems, expert testimony may be necessary to clarify the evidence’s evidence-generating process. However, this calls for the use of extremely specialised information that jurors and judges may not always understand or have access to. But the cost of hiring highly qualified specialists will raise the price of litigation even more, making it unaffordable.

(vi) Judicial Literacy

Since predictive coding is still in its infancy in Bangladesh, courts there may need assistance developing their ability to comprehend and assess AI technologies.^[16]

Due Process and Fair Trial Concerns

The Constitution of Bangladesh ensures the Right to a Fair Trial under the provision of Article 35.^[17] And this article is a fundamental right of a citizen of Bangladesh, which is being protected by the apex law of the country. Within the broader scope of this article, other fundamental rights like equality before law^[18], and equal protection of law^[19] can be included. This right is more likely to be threatened if AI techniques are used in forensic investigations without judicial approval. Because of this, erroneous AI-based forensic evidence might result in a wrongful conviction. Additionally, it can be challenging for the defence attorney to successfully cross-examine the testimony of an AI expert or to contest the proprietary algorithms and underlying source code, which would violate the right to a fair trial.

Recommendations and Reforms

Based on the study, the following recommendations can be suggested:-

(i) Legislative reform: To regulate the use of AI in evidence analysis, legislation specifically addressing the topic must be passed. Additionally, the present Evidence Act of 1872 shall undergo a major modification to specifically address the admissibility and dependability of AI-analysed evidence.

(ii) Ensuring transparency: Prior to being used in the criminal justice system, all AI technologies must pass a reasonable audit trial and have some explainability so that they may be successfully challenged in court.

(iii) Adapting AI admissibility standards: Bangladesh, like the United States, has to follow Daubert and Frye criteria, with a special emphasis on scientific testing and validation of AI and widespread acceptance in the scientific community despite its known and possible mistake rate.

(iv) Governance of AI ethics: An independent AI Ethics Board shall be formed to oversee the development and application of AI in the sensitive justice domain.

(v) Building capacity: Judges, prosecutors, defence attorneys, and forensic specialists shall all be required to complete mandatory training in AI literacy, data science, and the critical assessment of algorithmic evidence.

Conclusions

To conclude, the introduction of Artificial Intelligence in forensic evidence is a paradigm shift that the outdated Evidence Act of 1872 is not equipped to deal with. While traditional forensic evidence is verified through the traditional channel of the human expert’s testimony, algorithmic evidence operates as a black-boxed ‘black box,’ necessitating a primitive transformation of our legal calculus. Our examination shows that a new normative model is desperately needed to safeguard justice in this new era.

For the guarantee of evidentiary quality, the legal focus on Admissibility must change from verifying human qualifications to demanding rigorous Model Validation and Certification. Moreover, the consistency of AI analysed evidence is to be determined not by reference to subjective human consistency but by reference to verifiable Technical Objectivity, with transparency as to the algorithm, the input data, and the processing chain. Above all, the fundamental right to a fair trial requires that AI-analysed evidence must be fully Contestable. In the absence of legal and technical means of uncovering the basis beneath, a form of Explainable AI, due process principles are effectively thwarted, with machine efficiency favored over the accused’s right to confront the evidence. Only through a legislative response that establishes a tailored, adaptive framework can the nation achieve a lasting balance between technological advancement in the pursuit of crime and the constitutional imperative of justice and fairness.

Reference(s)

Statues

The Constitution of the People’s Republic of Bangladesh

The Code of Criminal Procedure 1898

The DNA Act 2014

The Evidence Act 1872

Cases

Kohls v. Ellison Case No. 24-cv-3754 (LMP/DLM)

Daubert v. Merrell Dow Pharmaceuticals 509 U.S. 579 (1993)

Frye v. United States, 293 F. 1013 (D.C. Cir. 1923)

^[1] Farsia Stoykova, ‘The Intersection of Forensic Science and Artificial Intelligence: Revolutionizing Crime Scene Analysis’ (2024) 9(6) JFM <https://www.hilarispublisher.com/open-access/the-intersection-of-forensic-science-and-artificial-intelligence-revolutionizing-crime-scene-analysis.pdf> accessed 6 October 2025

^[2] Zahurul Islam, ‘Strengthening Forensic Science in Bangladesh’s Criminal Justice System: Challenges, Opportunities, and Strategies for Improvement’ (2025) 9(16) IJRISS <https://dx.doi.org/10.47772/IJRISS.2024.916SCO0003> accessed 6 October 2025

^[3] Forensic evidence means evidence derived from the use of a field of science or the scientific method in order to investigate and prove crimes. Erin Murphy, ‘Forensic Evidence’ (2017) 3 RCJ

^[4] Manish Gupta, ‘How AI is Shaping the Future of Forensic Investigations’ (INNEFU, 12 July 2025) <https://innefu.com/how-ai-is-shaping-the-future-of-forensic-investigations/> accessed 7 October 2025

^[5] The Supreme Court outlined the following standards in Daubert for the trial court to use when deciding whether scientific knowledge would be admissible: whether the theory or technique has been tested, whether it has been published and reviewed by peers, whether it has a known error rate, whether there are established or upheld standards governing its use, and whether it has gained broad acceptance within the relevant scientific community. Daubert v. Merrell Dow Pharmaceuticals, Inc., 509 U.S. 579 (1993)

^[6] Whether an expert’s scientific knowledge “gained general acceptance in the particular field [to] which it belongs” is assessed under the Frye test. 1014 Frye. In other words, if the scientific method that an expert’s judgement is founded on is “generally accepted” as trustworthy in the relevant scientific community, then the view is acceptable. The Frye standard is used by Washington courts. Frye v. United States, 293 F. 1013 (D.C. Cir. 1923)

^[7] Musferat Mazrun Chowdhury and Ummay Habiba Naznin, ‘Forensic Evidence in the Criminal Justice System of Bangladesh: Laws and Practices’ (2022) 9(1) BUPJ p 84 – 85

^[8] Salwa Hoque, ‘Reforming AI Laws and Regulation in Bangladesh: Current Harms and Possible Future(s)’ (TECHGLOBAL INSTITUTE) <https://techglobalinstitute.com/research/reforming-ai-laws-and-regcancanulation-in-bangladesh-current-harms-and-possible-futures/> accessed 7 October 2025

^[9] Case No. 24-cv-3754 (LMP/DLM)

^[10] 509 U.S. 579 (1993)

^[11] Max M. Houck, ‘CSI/AI: The Potential for Artificial Intelligence in Forensic Science’ (ISHI, 29 October 2024) <https://www.ishinews.com/csi-ai-the-potential-for-artificial-intelligence-in-forensic-science/> accessed 8 October 2025

^[12] Supra Note to to 4

^[13] Evelina Gentry, ‘The Challenges of Integrating AI-Generated Evidence Into the Legal System’ (akerman, 12 June 2024) <https://www.akerman.com/en/perspectives/the-challenges-of-integrating-ai-generated-evidence-into-the-legal-system.html> accessed 8 October 2025

^[14] Supra Note 9

^[15] Kumar Shaswat Anand & Dr. Shailja Thakur, ‘Challenges and Limitations of AI in Forensic Science: A Critical Review’ (2025) 6 (1) IJRPR <https://ijrpr.com/uploads/V6ISSUE1/evidence-generatingIJRPR38264.pdf> accessed 8 October 2025

^[16] Trishita Chatterjee, ‘Admissibility of AI-Reviewed Digital Evidence in Legal Investigations’ (2025) 5(2) IJIRL p 2060

^[17] Article 35(1): No person shall be convicted of any offence except for violation of a law in force at the time of the commission of the act charged as an offence, nor be subjected to a penalty greater than, or different from, that which might have been inflicted under the law in force at the time of the commission of the offence. (3)  Every person accused of a criminal offence shall have the right to a speedy and public trial by an independent and impartial Court or tribunal established by law. The Constitution of the People’s Republic of Bangladesh

^[18] Article 27: All citizens are equal before law and are entitled to equal protection of law. The Constitution of the People’s Republic of Bangladesh

^[19] Article 31: To enjoy the protection of the law, and to be treated in accordance with law, and only in accordance with law, is the inalienable right of every citizen, wherever he may be, and of every other person for the time being within Bangladesh, and in particular no action detrimental to the life, liberty, body, reputation or property of any person shall be taken except in accordance with law. The Constitution of the People’s Republic of Bangladesh