The rapid rise in the use of generative artificial intelligence (AI) systems has streamlined information processing, drafting, and research, making these processes more efficient and less time-consuming. Professionals within the justice system are also taking up the use of AI tools for legal research, drafting, and case preparation worldwide. However, a notable concern remains regarding the fact that generative AI often produces false citations. If a legal decision is based on a non-existent or inaccurately represented authority, this could result in miscarriage of justice and erode public confidence in the judicial system.
In the legal context, such “hallucinations” may take the form of non-existent case laws, fabricated names of parties, false citations, misquoted or misrepresented judgments, and incorrect statutory provisions. The benefits of AI are well recognised. But if AI causes a legal error, who should be held responsible? This dilemma is the subject of legal and policy debate. AI is a tool; it is not a lawyer, judge, or an independent bearer of legal responsibility. Global practice in the courtroom is that responsibility remains with the human actor who places the information before a court or relies upon it in judicial reasoning.
Damien Charlotin’s AI Hallucination Cases Database has recorded some 1,934 incidents globally as of August 2026. The 2023 Mata v Avianca, Inc. case was the first recorded incident where fake cases produced by ChatGPT had been cited in a filing before a US court. The court fined the attorney $5,000 and held that attorneys have a gatekeeping duty to ensure the accuracy of legal authorities before citing them to a court. Thereafter, revealing AI-hallucinated citations have occurred in different jurisdictions, including in the United Kingdom, Australia, France, Canada, and even neighbouring India.
In the July 2026 Pooja Ramesh Singh v Jammu and Kashmir Bank Ltd & Anr case, the Indian apex court stated that any order passed on the basis of fake citations, even if in part, is “no decision in the eyes of law.” The court has also asked the Bar Council of India to frame the disciplinary rules for advocates for citing false authorities generated by AI. The landmark judgment has set a precedent for zero tolerance in India towards hallucinated citations and has reiterated the Indian judiciary’s commitment to authenticity and accountability. In the meantime, the Kerala and Gujarat high courts have come up with an AI policy for judicial officers regarding the responsible and limited use of AI in judicial functions.
In Bangladesh, the ICT Division has prepared a draft AI policy to develop AI responsibly. The Cabinet Division, meanwhile, recently issued cautionary directions not to put sensitive and confidential information into AI, while Bangladesh Bank has issued directives not to put confidential banking data into AI systems in order to prevent possible breaches. Chief Justice Zubayer Rahman Chowdhury has recently warned that AI must not be used to write court judgments, calling for laws and policies to clearly define where and to what extent the technology may be used in the judiciary. These developments show that various public institutions in Bangladesh are recognising both the opportunities and risks associated with generative AI.
Where does the judiciary stand in this emerging AI landscape? AI use is inevitable and will enter the judicial arena. Therefore, the need of the hour is a pragmatic directive on the use of AI in the judiciary. A key concern is low AI literacy among justice sector actors, including judges, lawyers, and judicial staff. A second problem is the availability and quality of digital legal databases. The absence of full digitisation and systematic indexing of legal databases increases the likelihood of legal citations being inaccurate or fabricated. Another problem is lack of an effective citation-verification mechanism particularly suited to the judiciary of Bangladesh.
In this regard, Bangladesh should form a judicial AI-use policy that is based on international experiences. The policy should clearly define areas where AI use should be restricted or prohibited. It must also prescribe appropriate sanctions for the misuse or irresponsible use of material produced by AI in the course of judicial proceedings. The Bangladesh Bar Council should frame rules for lawyers on the use of AI tools to ensure that they verify citations generated by AI and clarify accountability for submitting false legal authorities. Equally important is the education and training of judges, lawyers, and court staff in terms of AI use. In addition, Bangladesh should strive to create a reliable, authentic, and easily accessible digital legal database comprising up-to-date statutes, judgments, and other legal authorities, which would enable independent cross-checking of information generated by AI. Lastly, any material assistance received from AI systems in preparing petitions, pleas, or legal submissions must be disclosed transparently.
If properly guided and regulated, AI can become an innovative assistive tool to support legal research, case management, drafting, and other administrative functions, thereby potentially contributing to the reduction of case backlog. Bangladesh should therefore consider formulating a judicial AI policy that is best suited to its specific legal, institutional, and technological context.
Md. Ziaur Rahman is MPhil fellow at the Institute of Bangladesh Studies, University of Rajshahi. He can be reached at [email protected].
Views expressed in this article are the author's own.
Follow The Daily Star Opinion on Facebook for the latest opinions, commentaries, and analyses by experts and professionals. To contribute your article or letter to The Daily Star Opinion, see our guidelines for submission.
