Possible Friction Between the Supreme Court and High Courts

One of the more undiscussed challenges of this draft policy lies not in its content, but in its implementation across India's 25 High Courts each with its own administrative autonomy and varying appetite for technological adoption. The draft regulations themselves prescribe two different authorities: the Apex Body at the Supreme Court level and separate AI Committees at each High Court, creating a structure that is federal in design but untested in practice. To illustrate this friction concretely: the High Court of Gujarat had already released its own Policy on the Use of Artificial Intelligence in Judicial and Court Administration earlier this April 2026, well before the Supreme Court's draft was published. This is not the first time such tension has surfaced, the implementation of live streaming of court proceedings following the Supreme Court's landmark judgment in Swapnil Tripathi v. Supreme Court of India is a telling precedent, despite the court opening the door to live streaming, several High Courts still do not permit it and others have made it available only to litigants in select courts, reflecting precisely the kind of uneven adoption that may well repeat itself with AI. This raises a pointed constitutional question where two AI governance frameworks exist, which one prevails? The answer is not straightforward. While the Supreme Court is the Apex Court of India, High Courts are not subordinate to it in matters of their own administration and rule-making, drawing their authority directly from the Constitution. What remains to be seen is whether the Supreme Court's national framework will prevail in practice, or whether High Courts that have already invested in their own policies will chart a different course

Courts are only one part of the puzzle and the remaining haven't Decided

Court regulation by itself cannot regulate the full landscape of AI in legal practice, there are other critical stakeholders too and they remain conspicuously silent. The Government, both Central and State, is the single largest litigant in India and when one adds the legal work generated by state instrumentalities and government-owned companies, its approach to AI in legal work will shape the landscape as much as any court regulation. Yet not a single Central Government ministry has framed a policy on how AI may be used in legal work, leaving a significant portion of AI-assisted legal activity in a complete policy vacuum.

The second and more fundamental concern is one of constitutional propriety. The draft regulations grant courts wide powers to regulate the conduct of advocates, litigants and private entities in relation to AI use, including the power to pass such orders as the Court may deem fit against persons who submit fabricated or false AI-generated content. While the intent is understandable, this is far too wide a power to be assumed by courts through a self-prescribed regulatory framework. Such powers, particularly those affecting the rights and liabilities of citizens and legal professionals, ought properly to flow from legislation passed by Parliament. Consider a practical hypothetical, a litigant, as a matter of habit, drafts a legal notice using ChatGPT and sends it without informing his advocate; that notice then finds its way into the pleadings. In such a case, who is liable? Certainly not the advocate, who was entirely unaware of its AI-generated character and cannot reasonably be expected to have known and yet under the present draft, he cannot be fully absolved either, nor can he be faulted for refusing to take the brief on that ground alone. This gap is not a minor drafting oversight. It reflects a deeper failure to account for the reality that generative AI is already embedded in how litigants communicate, long before it reaches a courtroom. Without a proper legislative framework governing AI in legal practice, the regulation of its use will remain reactive rather than proactive and largely dependent on individual judges' discretion.

What's in it for private AI players?

There is both good news and bad news for private AI companies eyeing India's judicial technology market and unfortunately, the good news runs out rather quickly. The good news is that the courts do not intend to build their own AI systems from scratch, rather, they expressly contemplate engaging private players for permissible applications across court processes Regulation 46(1) and (2). The good news, however, ends precisely there. Regulation 46(1) requires every private entity to seek prior written approval from the Appropriate Authority before providing any service in connection with an AI system deployed in court processes which, in practical terms, means seeking up to 26 separate approvals: one from the Supreme Court's Apex Body and one from each of the 25 High Court AI Committees, with the technical and procedural requirements of each remaining largely unknown at this stage. Beyond the approval labyrinth, the more commercially damaging provision is Regulation 46(9), which states that where AI tools are developed using court data or court resources, "the Appropriate Authority shall ensure that the Court retains ownership of, or a perpetual royalty-free licence to, the resulting tool and its outputs" and that "no private entity shall claim exclusive intellectual property rights over tools developed primarily using judicial data or public resources." For most AI companies whose entire business model is built around training proprietary models on data and commercialising the resulting products, this is a significant potentially prohibitive condition. And lastly, one must not forget that red tapism in the Indian legal system is well and widely known. However, with new beginnings, let us just say things can always improve.

Author is a practicing Advocate before courts of Delhi and Haryana. Views are personal.