Why Judging with AI is not Legal

The Chief Justice (CJ) of Bangladesh remarked in the inaugural session of an event on August 29 that, ‘You cannot write a verdict with AI.’ ‘It is inappropriate, may be not illegal, but definitely not ethical.’ The Law Minister also recently opined that Artificial Intelligence (AI) can assist in judicial proceedings although judges cannot rely on it for analysing evidence. Judges do not formulate their decisions solely based on data, he added. Further, he commented that judges must meticulously examine evidence by applying their own concise, intellect, wisdom, foresight, and integrity to ensure justice. They have rightly recognised the issues relating to AI and wisely expressed their view that there should be a law to regulate the extent and manner of the use of AI. 

There is no room for disagreement with these thoughtful statements of them. In recent times, the vast expansion and regular application of AI tools, especially generative artificial intelligence and more specifically Large Language Model (LLM) chatbots such as ChatGPT have gained immense popularity due to their free availability. Therefore, a clear and comprehensive legislation is urgent to draw boundaries for arbitrary and irresponsible incorporation in daily use without properly knowing the functions and limitations of modern AI technology. In the absence of a standalone legislation, at least there needs to be a policy or guidelines for the judiciary. 

Article 107 of the Constitution of the People’s Republic of Bangladesh empowers the Supreme Court (SC) to make rules for the procedure and conduct of proceedings for both the divisions (i.e., Appellate Division and High Court Division) and subordinate courts. Hence, according to the Bangladesh Supreme Court (High Court Division) Rules, 1973 and the Bangladesh Supreme Court (Appellate Division) Rules, 1988, the apex court have the authority to promulgate rules regarding legal procedures. In several countries, the highest courts or judicial bodies have already adopted policies/practice directions/guidance/guidelines for judges, lawyers, litigants, and court staff as to the ethical, responsible and appropriate use of AI in their courts, including India (draft), Pakistan, Singapore, Hong Kong, Philippines (proposed), South Korea, Taiwan, Morocco, Ireland, Israil, Kenya (draft), South Africa (draft), the United Kingdom, the United States, Canada, Australia, Colombia, New Zealand, Belize, the Bahamas, Turks and Caicos Islands, Trinidad and Tobago, Jamaica, Italy, Spain, Paraguay and so on. Although there are questions about the effectiveness of these guidelines but at least these countries have a policy to regulate AI in the legal field. In addition, the regulatory bodies of lawyers in some countries have issued policies for them. 

Although it is understandable from the mentioned speech of the CJ that AI is also being used in the legal profession of Bangladesh. Like any other profession, this emerging technology can be utilised in the legal setor as well, but it is important to have an institutional stance on where, when, how and up to what extent it can be employed. 

Now, let me come to the CJ’s point of whether writing of judgments with AI is legal or illegal. In his speech, the CJ did not mean only writing judgments, instead he also meant adjudicating the disputes in the court. This is clear from the subsequent part of his speech where he says that he himself combines three techniques in determining disputes, e.g., facts (conflicting points); determination, application and interpretation of the law(s) and application of the conscience or morality; which in his words is significant and absent in AI.

It is true that there is no law currently in force in Bangladesh that specifically prohibits the application of AI to write judgments or judging court cases. But that alone does not mean it is not illegal to adjudicate or deliver judgments solely with AI. Until the existing procedural laws, constitutional principles and established legal norms are changed, it can be assumed that getting justice from a human judge is embedded in the right to fair trial. The Constitution of Bangladesh has clearly established the SC and vested judicial power in its judges. Therefore, only human judges can hear, interpret, adjudicate and deliver enforceable verdicts. Thus, a new legislation is required to empower AI to legalise trial by that.

Moreover, the independence of the judiciary is one of the cornerstones of the Constitution. If AI is integrated to judge, the independence of the judiciary will be definitely compromised. Further, judging by AI is against the principle of separation of powers enumerated in Article 22 of the Constitution. The challenge of judicial decision-making through AI also violates several fundamental principles of justice including the obligation to provide a reasoned decision, so that the aggrieved party can seek redress against such reasoning before an appellate court, and the exercise of judicial discretion and equity along with considering relevant context(s). For these reasons, the abstract concept of an AI or robot judge has already given rise to deep constitutional concerns in numerous jurisdictions. Presently, the use of AI in courts in liberal democracies is mainly confined as support tool to administrative tasks excluding core adjudication.

In conclusion, it can be said that even if it is not explicitly prohibited by law, judging or writing judgments with AI is in no way legal in our legal system due to its involvement with the basic structure of the constitution, the existing procedural laws, and established long legal tradition. As a result, any move towards a fully automated or AI-driven judicial decision-making requires necessary constitutional amendments and a radical change of current judicial processes.

Published in the Law and Our Rights Page of The Daily Star on 9 September 2026.

Published in the Weekly Shampratik Deshkal as Op-ed on 17 September on page 3.

Published in the Desh Rupantor as op-ed on 19 September on page 4.

Published in the Daily Ittefaq as Op-ed on 22 September 2026 on page 9.