Robot in the robe: How AI is reshaping the judiciary

For centuries, justice has been symbolized by a blindfolded figure holding a set of scales – an enduring representation of impartiality, reason and human judgment.

But there is now an unexpected guest at the courtroom door: Artificial Intelligence (AI). Unlike the traditional symbol of justice, AI does not come with a blindfold-or, for that matter, a conscience.

Across the world, courts and legal institutions are increasingly experimenting with AI for tasks ranging from administrative assistance and legal research to more advanced forms of decision-support.

So the question is no longer whether AI will enter the courtroom; it already has. The more interesting and far more consequential question is: How far should we let it go?

Finding an answer to this question assumes even greater significance in the context of Bangladesh’s judicial system, where the justice system faces a challenge that no amount of legal brilliance can easily solve: there are simply too many pending cases waiting for adjudication.

According to a statement by Law Minister Md. Asaduzzaman, the total number of pending cases across all levels of Bangladesh’s judiciary has crossed 46 lakh.

The numbers at the Supreme Court alone are striking: 38,713 cases were pending before the Appellate Division and 522,331 before the High Court Division.

Behind these figures are not merely case numbers, but people waiting for disputes to be resolved, rights to be determined and in many instances, lives to move forward.

This pressure is felt inside our courtrooms as well, where judges and judicial Benches must address a substantial number of matters within limited working hours.

The relentless pressure to hear and dispose of a considerable number of matters within a single working day not only affects the efficiency and quality of the judicial work but may also take a significant physical and mental toll on judges.

However, the problem is not a shortage of legal knowledge or commitment rather it is the sheer volume of material that modern litigation can produce.

A single case file may run into hundreds – or even thousands – of pages, containing pleadings, annexures, witness statements, previous orders and a small forest of supporting documents.

Reading everything carefully, finding the relevant passage, checking the procedural history and locating the applicable law is not merely intellectually demanding; it is also enormously time-consuming.

This is exactly where AI may have something useful to offer. Imagine a judicial assistant that never gets tired of searching through a 1,500-page case record, never loses a document under a pile of files and can look for a particular argument or precedent within seconds. That does not make the machine a judge.

It simply makes it a rather tireless but very efficient research assistant.
And that distinction is crucial.

Because AI should not be viewed as a substitute for judicial adjudication or human discretion.

Its potential value lies elsewhere: organising information, retrieving documents, summarising lengthy records, assisting with legal research and reducing repetitive administrative work.

If used properly, AI could help judges spend less time wrestling with information and more time doing what only a judge can ultimately do-assessing evidence, interpreting law, exercising discretion and making decisions for which a human being remains accountable.

Such use of AI in judicial systems is no longer merely theoretical, as courts and legal institutions in several jurisdictions have begun adopting AI-based tools for different aspects of judicial and administrative work.

In Egypt, for instance, AI-powered transcription tools have been developed as part of broader judicial digitalisation initiatives, enabling court proceedings to be transcribed more efficiently while reducing the burden of manual transcription.

Türkiye has similarly incorporated various AI-based functions into its e-Justice infrastructure, UYAP, including speech-to-text applications, legal-document summarisation and document classification in enforcement proceedings.

Our neighboring country,India has introduced SUVAS, a language-translation tool designed to make judicial decisions accessible in regional languages.

In the United Arab Emirates, Aisha, an AI-powered virtual assistant, has been used to support judges by analysing previous cases and providing relevant jurisprudential information.

However, Singapore has taken the idea in another direction. Its Small Claims Tribunal has explored AI to assist litigants through the claims process-helping identify necessary documents, facilitating accurate filing, providing information about possible outcomes and claim amounts.

The technology, in other words, is not necessarily sitting on the Bench; sometimes it is simply helping people find their way to it.

These examples reveal something important: there is no single model of “AI in the courtroom”. It can be a transcriptionist in one jurisdiction and a research assistant somewhere else.

Its role can be modest or if poorly regulated, potentially much more consequential. It is precisely at this point that a potential problem may arise.

AI systems are not neutral simply because they operate through mathematical models and code.

They learn from data and data can contain bias. They can produce errors with impressive confidence, obscure the reasoning behind their outputs or create serious privacy and data-security concerns.

There is also a deeper philosophical problem. Courts do not merely process information; they exercise judgment.

A judge may have to weigh competing rights, understand human circumstances or decide how a legal principle should apply to an unusual set of facts.

These are not simply search-and-sort exercises. They involve responsibility, discretion and accountability.

Perhaps, in the near future, we will find ourselves in courtrooms where humans and AI each perform the functions they are best suited to undertake.

Let AI search the mountain of documents, transcribe the proceedings and locate precedents.

But when the scales finally have to be balanced, the responsibility should remain firmly with the judge sitting on the Bench.

After all, the justice system may eventually become faster with AI, but faster justice is meaningful only if it remains justice.

(The writer is Advocate, Supreme Court of Bangladesh)