- with readers working within the Business & Consumer Services industries
- within Technology, Real Estate and Construction and Privacy topic(s)
- with Senior Company Executives, HR and Finance and Tax Executives
- in United States
Introduction
On June 3, 2026, the Supreme Court of India's Artificial Intelligence Committee published for public consultation its draft Regulations for Use of Artificial Intelligence in Courts, 2026 (“Draft Regulations”). The document represents a comprehensive attempt by the Supreme Court to govern the deployment of AI within a domain as constitutionally sensitive as adjudication. The deadline for stakeholder submissions is kept as June 20, 2026.
India's judicial system processes hundreds of millions of cases across its Supreme Court (“SC”), twenty-five High Courts, and thousands of subordinate courts and tribunals. AI adoption in this ecosystem is not speculative. Various tools for transcription, cause-list management and legal research are already in operational use. The Draft Regulations therefore are a regulatory response to a deployment already underway.
Principles Over Rules
Chapter II of the Draft Regulation articulates seventeen principles ranging from human primacy and judicial independence (Regulation 4) to a novel “Innovation over Restraint” clause (Regulation 17). This architecture borrows from the European Union's risk-based approach under the EU AI Act, 2024, which similarly drafted with feature of high-level normative commitments before descending into operational specifics1.
The principle of human primacy is unequivocal stated under Regulation 4 as AI systems should function solely in an assistive capacity and shall not supplant or compromise the independent exercise of judicial authority. By the way of this clause, SC tries to curb decisions taken solely on AI’s reliance. However, the Draft Regulations do not adequately address the gradual erosion of the independent human decision making in practice that is automation bias, the tendency of human decision-makers to over-rely on algorithmic outputs even when nominally retaining final authority. However, the Draft Regulations must also prescribe how this requirement should be substantively operationalized and not just merely declared.
Further, the contradiction between Regulation 16 in regard to presumption in favour of responsible AI adoption and Regulation 17 supporting innovation over restraint is analytically significant. The Draft Regulations effectively imposes an institutional need towards AI adoption that courts are expected to justify reasons for refusal or restriction on adoption which is a reversal of the conventional precautionary paradigm. While the intent is to prevent bureaucratic inertia from making use of advancing technological gains, it raises a legitimate concern as Indian courts still suffer from institutional resource constraints.
The Prohibited Uses Framework
The most consequential provisions of the Draft Regulations are the absolute prohibitions enumerated in Regulation 20. The regulations provide for nine categories of use that are rendered non-derogable and they cannot be relaxed even by the Apex Body or the Chief Justice under the general modification power conferred by Regulation 56. These include the prohibition on risk scoring for any purpose including evaluation of bail eligibility, recidivism and flight-risk assessment, the prohibition on AI-driven adjudication or sentencing without mandatory human in loop and the prohibition on undisclosed, opaque or unexplainable AI systems in proceedings that may affect personal liberty.
These prohibitions reflect an admirable alignment with emerging international consensus. The prohibition on risk-scoring tools is particularly notable given the documented harms caused by such systems in comparable jurisdictions, most prominently the COMPAS algorithm in the United States, which was found to exhibit racial bias in recidivism prediction2. The Draft Regulations categorical rejection of this category of tool for Indian courts is the correct regulatory response.
However, the prohibitions as drafted suffer from a structural deficiency as they lack an enforcement mechanism that is commensurate with their non-derogable status. Regulation 21 provides that violations shall be reported to the AI Secretariat, which places the matter before the AI Committee for such remedial measures as it deems appropriate. For prohibitions described as absolute and non-derogable, the remedial framework is extremely discretionary. The Draft Regulations are silent on whether violations attract disciplinary consequences for responsible officers, whether affected litigants have a right to challenge affected proceedings or whether convictions obtained in proceedings where prohibited AI was deployed are subject to review. This lacuna is not academic and it goes directly to the enforceability of the framework's most important commitments.
Institutional Design
Chapter IV creates a layered institutional architecture with an Apex Body at the Supreme Court level, AI Committees in each High Court, a Centre of Research and Excellence on AI (CoRE-AI), and AI Secretariats at each court. The design is comprehensive but the composition of the Apex Body raises questions about structural independence.
The Apex Body is chaired by a sitting Supreme Court Judge nominated by the Chief Justice of India and includes, among others, an officer of MeitY, advocates nominated by the Chief Justice and the National Judicial Academy AI Professor ex officio. This creates a conflict of interest in the dominance of judicial nominees in the body. The Draft Regulations do not address that the Apex Body that regulates AI in courts is itself appointed by the very court it seemingly regulates. Thus, the creation of Apex Body from a governance standpoint is structurally defective. The AI Secretariat framework also addresses an operationally important gap by vesting executive responsibility on District Judge-level officer for day-to-day AI governance at each court.
For comparative analysis, The UK's Centre for Data Ethics and Innovation3 operates at arm's length from both the government departments it advises and the regulators whose work it scrutinises. The Draft Regulations' CoRE-AI is placed "under the oversight" of the Apex Body which means the research body that should be providing independent analytical inputs is instead accountable to the body whose decisions it is supposed to inform. Such structurally broken accountability risks biasness where research conclusions can be shaped by the institutional preferences of the body that commissions and oversees them.
Transparency And Disclosure
Chapter V of the Draft Regulations provides for oversight, audits and incident management. The mandatory disclosure obligation under Regulation 43(3), requiring parties and their legal representatives to declare AI-assisted preparation of any document, pleading, or evidence through a prescribed certificate, is a significant step towards procedural integrity. Similarly, the obligation of courts to inform parties when AI tools materially assist in case management, operationalises the principle of transparency at the level of individual proceedings rather than only at the systemic level.
The AI Register mechanism and annual transparency report create a publicly accessible audit trail. This feature has been conspicuously absent from most administrative AI deployments in India up until now. The requirement that audit reports be recorded in the AI Register and that cybersecurity audits be conducted at intervals not exceeding one year, imposes a discipline of institutional accountability.
However, the provision that public dissemination of the AI Register shall be subject to data protection, confidentiality and cyber security might be used as a cover by many to evade transparency requirements. If the justification for limiting public access to the AI Register is allowed to be invoked at institutional discretion, the register risks becoming an internal document rather than the public accountability mechanism. The Draft Regulations could be modified to prescribe a minimum disclosure floor, specifying which categories of information must be publicly available regardless of confidentiality considerations.
Rigorous Safeguards For Private Sector Engagement
Chapter VI's provisions on procurement and private sector engagement are among the most practically consequential in the Draft Regulations. The mandatory contractual provisions enumerated in Regulation 46(4) includes prohibition on retraining court AI models on judicial data without Committee approval, on-premise or sovereign cloud requirements for sensitive data and mandatory indemnity clauses. This reflects a sophisticated understanding of the principal-agent risks inherent in AI procurement from private vendors.
Regulation 46(9) addresses a frequently overlooked intellectual property dimension that is where AI tools are developed using court data or court resources, the court will retain ownership or a perpetual royalty-free license, and no private entity may claim exclusive IP rights over tools built substantially on judicial data. This provision has significant downstream implications for the open-source availability of judicial AI tools.
Yet the framework contains a material omission. The Draft Regulations prescribe that audits of private vendor AI systems shall be conducted in-house under no circumstances sharing source code or algorithms with third parties outside court premises (Regulation 38(2)). While the rationale for this restriction is protecting source code confidentiality and cybersecurity, it effectively precludes independent external auditing of vendor-supplied systems. The result may be that audits are formally conducted by just ticking a compliance box without generating the accountability the framework intends.
Grievance Redressal Framework
Regulation 52 provides that a party harmed by a prohibited use of AI may file an application at the earliest opportunity to the court where the AI was used, with that court empowered to pass appropriate orders as it may deem fit. Regulation 53 preserves the right to seek remedies under other applicable laws.
This framework is inadequate for several reasons. First, requiring an aggrieved litigant to file an application before the same court where the violation occurred creates an obvious institutional conflict as the court is effectively asked to adjudicate the validity of its own process. Second, the phrase at the earliest opportunity imports a waiver-by-delay risk without any guidance on what constitutes timeliness in this context. Third, and most fundamentally, a litigant who is unaware that prohibited AI was used in their proceeding which is the most likely scenario given the opacity of many AI systems, it cannot invoke a remedy whose trigger depends on knowledge of the violation. The Draft Regulations do not address the gap between harm and awareness.
Conclusion
The Regulations for Use of Artificial Intelligence in Courts, 2026 is driven by principled regulatory initiative. The foundational commitment to human primacy, the categorical prohibition on risk scoring, the layered institutional architecture and the mandatory transparency obligations together represent a framework that compares favourably with AI governance regimes in comparable jurisdictions.
India's judiciary is among the most consequential institutions in the world's largest democracy. The AI systems that assist it will shape access to justice for over a billion people. The Draft Regulations represent a serious beginning towards accountability for unauthorized and harmful use of AI.
Footnotes
1. European Union. (2024, June 13). Regulation (EU) 2024/1689 Of The European Parliament And Of The Council
2. Pro Publica. (2016, May 23). How We Analyzed the COMPAS Recidivism Algorithm
3. Gov.UK. Centre for Data Ethics and Innovation (CDEI)
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
