There are projects in some jurisdictions – including China, Brazil, and Argentina – as to how these technologies could be employed in the judicial decision-making process to reduce case backlogs and enhance the delivery of justice. Contrarily, some countries – such as the UK, Ireland, New Zealand, and Canada – have taken an opposing stance to prevent the adoption of LLMs in core judicial functions, including determination and adjudication. Instead, they permit LLMs as support tools for judges and court staff to perform routine work like summarising text, speech writing, searching for information, sources, and precedents, enhancing clarity and linguistic flourish, translating, and transcribing.

However, such ‘innocent’ use of LLM tools in routine or secondary court functions has the potential to influence judicial cognition and thereby affect the core judicial decision-making process. Institutional permission and repetitive use of these tools may lead to several cognitive effects, such as automation bias, cognitive offloading, anchoring, and the framing effect. Automation bias leads to over-reliance on automated recommendations, while cognitive offloading reduces mental effort through external tools. Anchoring causes over-dependence on initial information, and the framing effect shows how decision-making shifts based on information presentation.

Judging involves reading the case materials, identifying the issues, and developing the reasons by themselves, whereas a preliminary recommendation produced by an LLM system has the potential to serve as a starting point for such cognitive involvement of judges, anchoring the point of departure. On the other hand, when a judge outsources the meaning of legal lexicon, a summary of long texts, or the identification of precedents to an LLM, they ultimately engage in cognitive offloading. This cognitive offloading can lead to a superficial review process where judges make only ‘surface-level’ scrutiny or modifications to LLM-generated text, bypassing the deep cognitive work required to fully understand the legal reasoning and verdict.

Therefore, simply relying on a human judge to use LLM systems in preparatory stages, or to audit the veracity of the outcomes of such systems subsequently in the judicial decision-making process, is potentially problematic. Though judges are trained to be impartial, cognitive and behavioural science demonstrates that no decision-maker is immune to systematic cognitive biases. Decades of behavioural research demonstrate that professional expertise does not eliminate susceptibility to anchoring, framing, or automation bias. Judges, like other professionals, work within cognitive constraints.

These cognitive defects are significant because human judges are regarded as the principal safeguard to check and control any type of bias and are also regarded as neutral decision-makers to ensure justice. In the circumstances of heavy judicial workloads, exhaustion, and in some cases institutional pressure to deliver timely decisions with limited resources, cognitive issues can arise. As a result, it can undermine the fundamental principles of fair trial, including judicial independence, the judicial duty to provide well-reasoned decisions, and ultimately the rule of law.

The inherent persuasive features of LLMs that randomly generate text with human-like reasoning and confidence, in spite of factual or legal accuracy, may make human judges vulnerable to this uncritical adoption of LLMs even as a support tool for judicial decision-making. This creates a deep tension with the fundamental and uncompressible judicial duty of independent reasoning, signifying that without precise safeguards and a deep understanding of this cognitive susceptibility, the integration of LLMs in adjudication, even as secondary tools, could unintentionally undermine the deliberative process it seeks to augment. Hence, in an increasingly technology-driven judiciary, it is critical to examine whether judges’ cognitive limitations prevent them from effectively guarding against aggravating this tension.

Domestic criminal law in countries like Bangladesh, India, and Pakistan permit this process to ensure trial of principal accused through the evidence of their accomplice, so that the trial is not hindered due to lack of evidence in complex and crucial cases. National criminal law usually sets conditions for such pardon including i) full, ii) true and iii) whole disclosure about the circumstances. If the approver does not fulfil the conditions, then the pardon will be revoked. The approver can be tried for the original offence for which they were pardoned, and their own confession can be used against them in that trial. Thus, the pardon initially offered to the approver does not provide blanket immunity.

Article 65 of the Rome Statute of the International Criminal Court (ICC) outlines a framework comparable to the provision of an approver. It governs proceedings based on an admission of guilt. Under this framework, the Chamber is not bound by the admission and must instead satisfy itself of several conditions. First, the accused must understand the nature and consequences of the admission. Second, the admission must be made voluntarily following sufficient consultation with defence counsel. Finally, the admission must also be supported by the facts of the case. This assessment is based on the charges, evidence presented by the prosecutor, and any other materials provided by the accused.

Nevertheless, it is clear in the Rome Statute that the accused remains accused and does not turn into a prosecution witness like in domestic law, and the Chamber may convict the accused even if the accused satisfies the above requirements. Hence, there is no scope for a predetermined or lenient sentence for an admission under Article 78 of the Rome Statute. This is because of the jus cogens character of international crimes. Consequently, pardon for such crimes should not merely be a domestic policy choice.

Additionally, the Guidelines for Agreements Regarding Admission of Guilt adopted by the ICC heavily emphasize the Court’s independent duty to establish the truth. Judges must examine not just the agreement between the parties, but also other evidence presented by the prosecutor and any other evidence, which can include victim representations and other sources, to ensure the facts are complete. Besides, the Guidelines explicitly recognize the need for victims to be informed of the proceedings and to be able to present their views and concerns, ensuring the process does not overlook their interests.

In contrast, the statutes and rules of the International Criminal Tribunals for the former Yugoslavia (ICTY) and Rwanda (ICTR) contained provisions for formal plea agreements. An accused could plead guilty to specific charges, often after negotiations with the prosecutor, who might agree to drop other counts or recommend a sentence. Although the Trial Chamber was not bound by the agreement. Judges had to verify that the plea was voluntary, informed, and unequivocal and that a sufficient factual basis for the crimes existed. In these cases, a convicted person received a sentence determined by the judges, with a guilty plea being a significant mitigating factor, often leading to a substantially reduced term.

Nevertheless, the primary goals of international criminal law and tribunals are not just to punish, but to establish a historical record of atrocities, promote reconciliation, deter future crimes, whereas a unilateral pardon could undermine these goals by appearing to offer impunity. Sentences are meant to reflect the individual’s guilt and the gravity of the crime, while a pardon would circumvent this carefully calibrated judicial process

Bangladesh witnessed wilful full blackout of internet for 5 days consecutively and 11 days for mobile internet across the country during the ‘Quota Reform Movement’ to quell the July-August mass uprising in 2024. In addition, the then government had sweeping powers to shutdown or slowdown internet on numerous occasions without providing any valid legal cause. As a result, there is a growing demand for an alternative, modern, and resilient connectivity solution beyond traditional internet service infrastructure.

Despite that, there is scope for wiretapping and shutting down internet under the draft telecommunication law and the Regulatory and Licensing Guidelines for Non-Geostationary Orbit (NGSO) Satellite Services Operator in Bangladesh prepared by the Bangladesh Telecommunication Regulatory Commission.

The United Nations (UN) recognised internet access as a fundamental human right in 2016. Subsequently, the UN Human Rights Council (UNHRC) adopted a non-binding resolution titled the promotion, protection and enjoyment of human rights on the Internet in 2021. Along with 42 other countries, Bangladesh also voted in favour of this resolution. Although the resolution is non-binding but it’s a commitment from the signing state parties to uphold the value of the resolution.

The ramifications of internet shutdowns on human rights can be severe and extensive. State censorship and surveillance through internet and telecommunications have detrimental impact on the human right to seek, receive, and disseminate information. Hence, international human rights instruments and standards uniformly reject internet shutdowns as a means of information control. Restricting access via a partial or complete internet blackout is characterised as indiscriminate and disproportionate restrictions. Any restrictions on information dissemination systems must comply with the three-part tests for restrictions on freedom of expression i.e. legality, legitimacy and proportionality. The responsibility to justify the tests is with the State attempting to limit the rights.

The legality principle provides any restriction must be provided by law and perhaps that is why Bangladesh is also interested to keep a scope in the law. However, the law must be concisely drafted, and it must be accessible to the public. The law should mandate strict oversight over the scope of such restriction and facilitate forum for judicial review to prevent any misuse. Additionally, the law must clearly specify the extent of discretion granted to public bodies and the manner of executing the discretion. Therefore, the proposed law should fulfil these criteria, as if a law fails to explicitly explain the relevant circumstances and conditions for internet shutdowns is detrimental for human rights.

The legitimacy rule stipulates any limitation on freedom of expression and other rights safeguarded by international law must aim to achieve a reasonable objective consistent with the identified grounds and be essential for attaining that objective. When states implement internet shutdowns or interrupt access to communication platforms, the legal basis for their actions is mostly undefined. A 2022 report of the OHCHR on internet shutdowns stated that the official justification for the shutdowns was unidentified in 228 instances across 55 countries from 2016 to 2021.

Finally, the proportionality criteria prescribe a restriction must be the least intrusive alternative and must not undermine the substance of the right. An appropriate evaluation of the efficacy of the restrictions and the danger of excessive blocking is to be made. This evaluation should ascertain whether the restriction may result in a disproportionate prohibition of access to internet or specific content, and whether it represents the least restrictive method to achieve the stated lawful objective.

Nonetheless, international law does not prohibit internet shutdowns completely; states may cut off internet during an actual state of emergency, or for national security and public order, but under strictly necessary circumstances and the least restrictive means. Most of the time internet shutdowns are indiscriminate, and disproportionate, cutting off citizens from access to information, education, emergency services, and more. As such, they typically fail the proportionality and legitimacy tests due to their arbitrary and extensive effects.

After the sacrifice of students and people, this is high time for a more actionable, realistic, and effective way of securing internet access in Bangladesh. Consequently, urgent review of the draft law and revision of the already issued guideline are essential to ensure they meet the above criteria to uphold human rights and free flow of information.

Reasons behind this Judgment

Like other Asian countries, many Bangladeshi families prefer to have male children for various economic, social, cultural and religious reasons. Consequently, the probability of termination of the foetus is higher by disclosing the sex before birth as female, although abortion is prohibited in the country.

There are multilayer risks to revealing the sex of an unborn offspring in a society like Bangladesh including forceful sex-selective abortion. Expectant mothers face potential physical and mental abuse, abandonment, divorce, or even death by male partners and in-laws.

Can the Verdict Prevent Gender Discrimination in Bangladesh?

This judicial pronouncement may be regarded as an easy solution but may not be the efficacious or humane attitude to eliminate gender inequality. This initiative would not be fruitful at enhancing gender equity without emphasizing other initiatives to change societal norms and root causes, and thereby reducing the demand for sex identification and forceful abortion.

Additionally, it cannot prevent the risk of maltreatment towards unwanted girls and their mothers in the household. The higher ratio of male to female citizens was attributed to the killing or neglect of female infants, even before the evolution of technologies to detect the sex in womb.

On the contrary, the implementation of law is a common challenge in Bangladesh and perhaps that is why the court tightened the scope for prenatal sex detection despite having a strict anti-abortion law. The underprivileged will possibly resort to unsafe and costly illicit means for sex identification due to the lack of monitoring of enforcement whereas the affluent sections of the country may travel to where neither abortion nor sex selection are illegal. Nonetheless, the effectiveness of the guidelines is also dubious as it only regards the non-compliance of its provisions by medical practitioners as a professional ethical issue only. Hence, the implementation of the guidelines will be really challenging without any binding legal framework.

Way Forward

In the pursuit of a more equitable society, it is imperative to move beyond mere prohibitions and encourage a comprehensive approach that tackles the root causes behind gender inequality in the country. Promoting gender equality in all spheres of the society in Bangladesh and prioritisation of women’s participation in economic engagement, particularly in employment and access to assets such as real estate and financial assistance are essential to enhance their engagement in the State economy. Acknowledging women’s unpaid domestic labour in the national GDP and advocating for equitable home obligations could be a vital measure. Moreover, facilitating women entrepreneurs and enhancing infrastructure will promote economic inclusion and eliminate son preference from the parents.

Strengthening the enforcement of law against gender-based violence, reform of judicial systems to ensure just disposal, enlarging social welfare programmes and services for women, and the expansion of women’s participation in political process (including leadership and decision-making) are key to ensure equality in the country. Underprivileged classes including Dalits, hijras, fisher-folk, sex workers, and ethnic and religious minority women must be equally integrated into policymaking and State development frameworks.

Conclusion

Simply discouraging foetal sex identification through executive guidelines is an inadequate attempt that does not confront the entrenched social and economic disparities fuelling gender inequality in Bangladesh. Limiting access to bodily information to prevent anticipated abortions harms individual autonomy as well. Thus, a blanket policy despite having rigorous anti-abortion law also affects people who would never consider abortion. In the absence of legally mandated enforcement and comprehensive socio-economic reforms, the verdict will predominantly be symbolic and ineffectual.

AI emerges as a beneficial tool for lowering the effort and cost needed to examine the documents, determine and apply appropriate provisions of law to a given fact, and increase accuracy by generating predictions. As a result, judging by AI arguably has the potential to be fairer and more neutral than human judges.  Therefore, some regard it as a cheap, fast, and scalable alternative. Human judges are by nature expensive as they have prepared for years, take time to adjudicate, retire, and are also limited in number, while AI systems can work more than 8 hours in a day tirelessly, do not take time off, and receive zero wages.

Dangers of Artificial Judicial Decision-Making

However, the accountability of judges can be compromised and weakened through the integration of AI tools in the decision-making process. This is because judges are likely incapable of delivering or clarifying the reasons an AI system produced while making a decision where the vendor’s software they rely on does not provide detailed information about its functionality and is not transparent.

Often, the functionality of an AI system is not revealed for the sake of operational secrecy or to protect trade secrets or privacy of personal information in training data. On the contrary, providing just and reasonable cause in judgments is one of the fundamental principles of justice.

AI systems are often trained with public source data which are not always authentic. Moreover, they do not have the self-capacity to evaluate and adapt to the social changes of the time. Furthermore, AI tools are not capable of applying discretion like human judges in specified circumstances. This may create injustice in many cases, as tech tools are not eligible to evaluate each case with appropriate considerations. Some cases may require progressive attempts and favour from the court to bring the marginalized into the mainstream, which algorithms cannot do on a case-by-case basis.

The quality of output given by AI is also questionable, as it relies on a vast past database, which may lead to inaccurate, incomplete, misleading, or out-of-date outcomes. Hence, there is a high risk that algorithmic judges would replicate the previous mistakes, discrimination, and bias of former cases. Discrimination may also result from the selective use of technology by human judges, and the susceptibility of algorithms to different cognitive biases.

Another dimension of integrating algorithms into judicial decision-making is accurately translating the legislation into codes, commands, and functions that a computer program can understand. Also, generative AI often produces fictitious case references (often called AI hallucination), incorrect interpretations, or quote overruled decisions.  This needs to be addressed before deploying AI in the decision-making practice, especially to preserve the right to fair trial.

Consequently, incorporating Algorithmic Decision Making (ADM) in judicial decision-making processes may vitiate core judicial values like the fairness of justice, diversity, equality before law, right to equal protection of law, and the right to privacy.

Public Concerns

Moreover, how can the court guarantee the security of clients’ privileged data when shared with AI algorithms? How does the public keep confidence that the information shared with AI is securely protected? Additionally, there is a risk that AI algorithms could be hacked or manipulated, which could lead to wrongful convictions. There is also the potential for power imbalance between parties to a lawsuit as the richer group may have more affordability and control over the use of AI systems than the marginalized. Therefore, public trust in the judiciary may be diminished as the public may not trust AI judges to make fair and impartial decisions.

What Can be Done

Adequate training on the functions and negative impacts of AI in the judiciary is essential for judges. Institutional oversight should be urgently employed to ensure that AI is used responsibly and cautions have been taken to mitigate the risks, as regular auditing by a superior authority can create an extra shield against the irresponsible and unethical use of AI.