LegRAA, Digital Courts 2.1, and the Emerging AI Layer of Indian Courts

LegRAA (Legal Research Analysis Assistant) is an AI tool built under the eCourts Phase III project by the National Informatics Centre's Artificial Intelligence Division together with the Centre of Excellence for eCourts (CoEE) in Pune. It is designed to help judges - not advocates or the public - search and analyze legal documents, taking pleadings and exhibits as input and retrieving and summarizing relevant precedent from a corpus of roughly 36,000 Supreme Court judgments.

Digital Courts 2.1 is a separate, complementary initiative aimed at giving judges and judicial officers a single, paperless interface for case-related information and tasks. It incorporates SHRUTI, a speech-to-text tool that lets judges dictate orders instead of typing them, and PANINI, a translation tool for moving legal text between Indian languages.

Both tools are being developed and piloted under the broader eCourts Mission Mode Project - the same initiative behind eCourts Phase III's digitization push (see our companion article on Phase III). The AI-specific component of that project has an earmarked outlay of roughly Rs 53.57 crore under its Future Technological Advancements component.

It's worth being precise about what these tools are: government-built systems, developed by NIC and the eCourts Centre of Excellence, intended for use by judges within the court system itself. They are not products of any private legal-tech vendor, and LegalDreams has no partnership, integration, or affiliation with LegRAA, Digital Courts 2.1, or the eCourts Mission Mode Project - nothing in this article should be read to suggest otherwise.

The distinction matters practically too: LegRAA searching a corpus of Supreme Court judgments for a judge's use is a different function from a private research tool an advocate might use to prepare a filing, even though both involve AI-assisted legal research.

As of this writing, LegRAA and Digital Courts 2.1 are described in official and independent reporting as being in pilot testing, not universal deployment across every court. That distinction matters for advocates: a tool being piloted at some courts doesn't mean it's in use at the specific court where a given matter is pending, and its eventual rollout timeline, scope, and final feature set may still change.

The judiciary building its own AI research tools - with a defined corpus, a defined user (judges), and a defined task (research and summarization, not adjudication) - is a useful reference point for how AI can be scoped responsibly in a legal context generally. It's a similar boundary to the one described in the Supreme Court's own draft AI Regulations for Courts, 2026 (see our companion article): AI assists with a bounded, verifiable task; the person using it remains responsible for what it produces.