INTRODUCTION
India`s judiciary is trying to figure out how to use Artificial Intelligence tools in its work and decision making. On June 3rd, 2026 the Supreme Court of India put out a draft of the Regulations for Use of Artificial Intelligence in Courts 2026. They want to hear what the public and lawyers think about it. The goal of this regulation is to make things faster with technology but to make sure the rights of the people involved in court cases are protected and that humans are always in charge as stated in Regulation 4.
The new draft regulations say that humans must always be involved in the process, which is called the ‘Human-in-the-loop’ requirement. Regulation 3(1)(zb) says that a human must check every step that is automated and that no task that requires judgment or discretion can be given to technology alone. This paper will look at how this new law came about what the new draft regulations say about using Artificial Intelligence and what the courts have recently said about using automated tools.
Artificial Intelligence tools, like Large Language Models and predictive algorithms are becoming more popular quickly but the courts are having trouble keeping up. These tools can help people work with documents understand languages and do research faster but they can also cause problems with the way courts work. By making laws about how to use these tools the Supreme Court of India wants to protect the integrity of the law system and make sure Artificial Intelligence is used in a way that helps the courts and the people and that Artificial Intelligence tools do not undermine the procedures. The use of Artificial Intelligence is an issue and the Supreme Court of India is trying to find a balance between using Artificial Intelligence to make things faster and protecting the rights of the people involved in court cases and this is why they are releasing the Regulations for Use of Artificial Intelligence in Courts 2026 to get feedback from the public and the legal profession, on the use of Artificial Intelligence.
1. The Judicial Catalysts – From “Phantom Precedents” to “Misconduct”
The draft regulations were created due to the rising use of ‘Generative AI’ tools in preparing court filings and trials. These tools, when prompted, generate references to false or distorted legal authorities. The issue started when trial courts and High Courts began using these tools to find and create references to non-existent precedents.
In this context, the Supreme Court addressed the concerns raised in a public interest litigation filed by Pooja Ramesh Singh. This litigation challenged the orders of a trial court in the case of Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. (July 2026). In Singh v. J&K Bank Ltd., a Bench of PS Narasimha and Alok Aradhe overturned the tribunal’s orders, which had cited AI-generated citations for support. The Bench stated that relying on AI tools to generate decisions “pollutes and poisons the stream of justice,” warning that unrestricted use of such tools could lead to “gasping in gas chambers of artificial intelligence.”
The Court also established a policy that presenting and relying on a judgment with false references would lead to disciplinary action and be considered misconduct.
2. Permitted Functions and Prohibited Acts – AI in Court
A dichotomy between the permissible and prohibited functions was made explicit by the Supreme Court in the 2026 Draft Regulations for Use of Artificial Intelligence in Courts. The “Administrative” functions listed under Regulation 19 are as follows:
1. Conversion of speech to text
2. Translation of documents in vernacular language
3. Classification of filings
4. Classification of documents
5. Searching of precedents on case law
However, the following adjudicatory functions are prohibited under Regulation 20 by the use of AI tools:
1. Risk assessment of bail application
2. Profile of witness credibility
3. Outcome determination or automated dispute resolution
4. Bail/sentencing determination
5. Behavioral pattern assessment of witness
Through this regulation, the Supreme Court ensures that the Right to Life and Personal Liberty of its citizens is safeguarded. In Anita Kushwaha v. Pushap Sudan (2016), this Court held that a fair and objective hearing forms an essential element of the Right to Life, as provided under Article 21 of the Indian Constitution. Automated determination/prediction of the result of the dispute would undermine the equality of hearing and introduce an element of “unpredictable discrimination”.
3. Regulation 8 – Strict Personal Liability
The requirement of Human-in-the-loop is given a narrow interpretation under the 2026 Draft Regulations. Regulation 8 explicitly states that no one will be relieved of his/her responsibility whether it is the person involved or an official appointed by him/her. The legal practitioner cannot use any of the following arguments in Regulation 8:
1. Mistake of computer programme or machine learning model;
2. Defense of “Black Box”, as provided for in Regulation 3(n) of the 2026 Draft Regulations; and
3. Hallucinations of the Generative AI, as provided for in Regulation 3(z) of the 2026 Draft Regulations.
In addition to this, Regulation 43(6) provides that when in a pleading, the inclusion of forged documents is made, then the presumption will be that the inclusion has been made with the knowledge and consent of the counsel of the petitioner. Therefore, in a case where there is an objection to an inclusion of AI-generated citations, then the burden of proving that the counsel was not aware of and did not have an opportunity to review the content will be on him/her.
4. Junior Advocates and Junior Officers – A Duty of Due Diligence
The ‘Human-in-the-loop’ rule is a transfer of duties and imposes another duty on advocates and officers to scrutinize the work done by their juniors. In India, senior advocates usually delegate some of their work to their juniors. Further, senior judges also delegate the task of filing to junior officers of the Court. It is obvious that junior officers and advocates make use of different AI instruments like legal chatbots for preparation of notes, conducting research and drafting documents. Since, LLMs used in the instrument only make predictions of output based on statistical probability and don’t apply any reasoning or conscious knowledge of the task, there is a possibility of errors even at the stage of delegation itself. The responsibility to check the assumptions made while making the document lies with the officer according to Rule 43.
Moreover, the Supreme Court, in its directives to the Bar Council of India (BCI), has laid down that, except for the signature of the advocate to authenticate the correctness of the document, the liability of the default position will lie on the signing advocate. In Singh v. J&K Bank Ltd., the Bench issued a caution that the “unbridled and unchecked use of AI technologies” in drafting court documents “contaminates and poisons the very source of justice.”
5. Constitutional Implications and the Future of Automated Adjudication
The use of Artificial Intelligence (AI) in the Indian context could trigger various constitutional controversies concerning the rule of law, judicial review, and other freedoms under Articles 14, 19, and 21 of the Constitution.
5.1 Article 14 and Non-Discrimination
The mandate of Article 14 precludes the State from engaging in any activity that would be considered discriminatory or whimsical in nature. Where there are individual acts involved in the implementation of the laws and equitable circumstances are to be considered, then use of software may affect the judicial processes that involve balancing different social considerations. In case a computer program is capable of giving a lighter sentence to a similar crime, it will amount to a violation of the mandate of equality before law.
5.2 Article 21 and Reason for Judgment
The provision of Article 21 that litigants have the right to a reasoned judgment means that an order cannot be passed without giving a proper justification for it. The litigant has the fundamental right to know the reason why the decision was passed. The judge should, therefore, not give reasons for his judgment through some randomly generated algorithm. HITL approach and need of reasoning of the orders passed by the judge in Indian courts are very important here.
6. Interplay with the Digital Personal Data Protection (DPDP) Act, 2023 and Data Governance
The integration of AI tools into the courtroom system is not separable from the Data Privacy law introduced by the government of India in the form of the Digital Personal Data Protection (DPDP) Act of 2023. Lawyers and court registries who upload their clients’ cases, witnesses’ statements, medical files and/or confidential corporation details to a third-party AI app in order to get translations or summaries are likely to infringe on their clients’ data privacy rights.
According to the 2023 Digital Personal Data Protection (DPDP) Act, court submissions comprise “personal data” of the parties concerned and the service providers of third parties. Uploading of unredacted court case files, medical files and other data files to an enterprise server of the commercial cloud-based Large Language Models (LLMs) exposes the personal data of the persons concerned to being scraped, mined, processed and exchanged by the LLM hosting the AI application. The 2026 AI regulations mandate the usage of DPDP Act-compliant AI systems by mandating the deployment of LLMs hosted by enterprise servers making zero-retention data possible. The Human-in-the-Loop process for using AI necessitates manual redaction of personal information from the documents submitted for uploading to the LLMs for generating translations, indexing and/or summaries.
7. Comparative Perspective:
Lessons from the EU AI Act and the United States Judicial Practices. India is not the only country that regulates the use of artificial intelligence in court procedures since there is another similar phenomenon in the European Union Artificial Intelligence Act. In particular, the act regulates the advanced use of artificial intelligence in the judicial sector as high-risk applications that require strict regulations to ensure human oversight, transparency, risk management, and accuracy. Particularly important in the regulation is human oversight to avoid “automation bias.
”This very problem has been identified by many US courts adopting judicial orders to tackle the emergence of generative AI. Particularly interesting is the United States District Court for the Northern District of Texas, which has imposed a standing order requiring all lawyers to certify that any document created with the help of artificial intelligence generated technology has been reviewed and approved by a human lawyer. There have been similar orders adopted by other federal district courts, and it can be predicted that there will be more such cases.
However, similar principles of professional integrity and personal accountability are reflected in the 2026 Draft Rules, which additionally stipulate certain issues in a much more elaborated way because they strictly prohibit the application of AI technology to the performance of basic judicial processes, while at the same time making judges, lawyers and other staff members accountable for the verification of the accuracy of information generated through the use of AI technology. Hence, the principle of “Human-in-the-Loop” appears to be a means of preventing the decline of professional accountability.
CONCLUSION
The ‘Human-in-the-loop’ requirement has now been made mandatory through Regulation 3(1)(zb) of the Supreme Court Regulations for Use of Artificial Intelligence in Courts, 2026. It is made evident through the 2026 Regulations that while such tools may be increasingly being used, such technology cannot override human instinct and conscience. The 2026 regulations have ensured that the new set of regulations does away with the “Black Box” and “AI Hallucination” defences as per Regulation 8, to make sure that the ultimate responsibility for making judgments, exercising discretion, and making decisions always lies with the human professional. The ‘Human-in-the-loop’ mandate ensures that the primacy of the human mind in judicial and quasi-judicial matters is not compromised, and due process of law is followed.
THIS ARTICLE IS WRITTEN BY JAYNAM SHAH FROM JITENDRA CHAUHAN COLLEGE OF LAW
REFERENCES
Regulation 3(1)(zb), Draft Regulations for Use of Artificial Intelligence in Courts, 2026
Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. and Anr., Civil Appeal no. 11950 OF 2025 SLP, Judgment dated July 2, 2026 (Supreme Court of India, as per P.S. Narasimha and Alok Aradhe, ).
Trial Court Uses AI-Made Judgments; Supreme Court Says “Misconduct, Legal Consequence Shall Follow”, LawBeat (March 2, 2026) LawBeat News.
Anita Kushwaha v. Pushap Sudan, (2016) 8 SCC 509 (Holding access to justice as a fundamental right under Article 21 of the Constitution of India).
