The e-Courts foundation
The groundwork was digital infrastructure, not AI. The e-Courts Mission Mode Project, running through successive phases since the mid-2000s, digitised case filing, case status tracking, and cause lists across thousands of courts. This unglamorous but essential work created the structured data — judgments, orders, case metadata — that any later AI tool would need to function. Without e-Courts, there would have been no usable corpus for AI to work with at all.
SUVAS: translation as the first AI application
The Supreme Court's AI-assisted translation tool, SUVAS, was among the judiciary's first visible AI deployments, built to translate judgments between English and Indian languages. Its early and sustained success — translating tens of thousands of judgments and expanding to support around 19 languages — established a template the judiciary would return to: AI as an assistive tool solving a defined, bounded problem, with human review built in rather than treated as optional.
SUPACE: research assistance for judges
The next major step was SUPACE, built to help judges and their research staff extract facts and map precedent from voluminous case files. Where SUVAS solved a language problem, SUPACE addressed an information-management problem — turning the sheer bulk of Indian litigation into something a judge could process faster, again without ever suggesting an outcome.
High Court policies and the first guardrails
As generative AI tools became widely available from around 2023 onward, individual High Courts began issuing their own policies rather than waiting for a national framework. Kerala's 2025 policy, followed by Gujarat, Punjab and Haryana, and others, restricted AI use in judicial decision-making and warned against feeding case data into public chatbots — an early, decentralised response to a fast-moving technology, built court by court rather than top-down.
2026: a national framework begins to take shape
The Supreme Court's Draft Regulations for the Use of Artificial Intelligence in Courts, 2026, mark the first attempt at a comprehensive, judiciary-wide framework — formally defining AI hallucination, drawing a bright line against AI deciding cases, and signalling disclosure requirements to come. Read against the decade before it, the draft looks less like a reaction to a sudden crisis and more like the natural next step of an institution that has been building and testing AI tools deliberately since long before the current public conversation began.
What the arc tells advocates
The consistent thread across every phase — SUVAS, SUPACE, the High Court policies, and now the national draft — is that the judiciary has never treated AI as a decision-maker. It has treated it as an increasingly capable assistant, with human judgment retained at every consequential step. That consistency is a useful compass for advocates building their own AI practice: the boundary the courts have held for a decade is the boundary worth adopting in your own chambers.
The takeaway
AI in Indian courts is not a sudden disruption; it is the latest stage of a careful, decade-long institutional project. Advocates who understand that history will recognise the current draft regulations not as an unprecedented clampdown, but as a codification of a principle the judiciary has quietly held since SUVAS's earliest days: assist, never decide. Platforms built for advocates, like LawWorld, are designed around that same enduring principle.
This article is for general information and does not constitute legal advice.
