Legal advice for founders of startups and micro, small and medium enterprises (MSMEs) in India remains complicated due to the decentralization of statutory rules, judicial decisions, and regulatory instructions across many sources. While the Indian startup ecosystem has expanded to over 100,000 acknowledged ventures, founders continue to face problems in legal issues around incorporation, equity structuring, compliance, and governance. Large language models often produce assertive yet unsubstantiated answers, with hallucination rates in commercial legal AI reported at up to 17-33%. This paper introduces JurisGPT, a domain-specific legal research assistant built on a retrieval-augmented generation (RAG) pipeline over a curated Indian legal corpus including statutory provisions of the Companies Act 2013, case briefs, a founder-agreement clause repository, and compliance information, served via a Next.js chat application backed by FastAPI. The system produces citation-grounded responses with confidence indicators, source-aware constraints, and follow-up suggestions. Across 120 startup and corporate law queries spanning six categories (annotator agreement: Fleiss kappa = 0.81), JurisGPT achieves Recall@5 = 0.78 and nDCG@5 = 0.72; groundedness improves from 34% (plain LLM) to 89%, and hallucinations drop from 42% to 7%. On a 5-point Likert scale, human annotators rate JurisGPT more trustworthy (4.5 vs. 2.1) and more useful for citation (4.3 vs. 1.0) than the non-retrieval baseline.
Bruhadev et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: