LawGPT: Connects Users with Justice via AI

Abstract

Access to justice continues to be limited, especially for individuals with limited awareness of legal procedures and limited access to professional guidance. Traditional methods of connecting with lawyers involve time-consuming processes, high costs, and geographical limitations. Our project, Law GPT, introduces an AI-powered legal assistance system designed to bridge the gap between users and justice. Law GPT allows users to register, log in, and submit legal queries in natural language. These queries are processed by an Embedding Model and stored in a Vector Database for semantic search. Relevant legal documents and precedents are retrieved and passed to a Large Language Model (LLM) that generates accurate, context-aware responses. Users can also connect with verified lawyers for professional assistance. The system reduces barriers to justice by: Providing instant AI-based legal guidance. Offering document search across case laws, acts, and precedents. Enabling connections with legal professionals through the platform. The system is designed as a scalable and secure framework for cost-effective legal assistance for improving public access to legal resources.

Introduction

This study introduces LawGPT, an AI-powered legal assistant designed to improve access to justice in India by helping citizens understand legal rights, procedures, and documentation through natural-language conversations. Many people face barriers such as complex legal language, high legal costs, and lack of awareness of legal rights. LawGPT addresses these challenges by providing legal guidance, document drafting assistance, and referrals to verified legal professionals when necessary.

Unlike traditional legal databases and search tools, LawGPT combines retrieval, legal reasoning, explainability, privacy, and multilingual support. It is specifically tailored for the Indian legal system and supports regional languages, making legal information more accessible to diverse populations.

Objectives and Motivation

The system aims to:

• Simplify legal information and procedures.
• Provide conversational legal assistance.
• Generate preliminary legal documents.
• Improve transparency through source citations.
• Connect users with qualified lawyers when needed.
• Ensure privacy and compliance with data protection regulations.

Literature and Related Work

Previous legal AI systems such as ChatLaw, CHRExpert, Aalap, and Legal Assist AI demonstrated the effectiveness of large language models (LLMs) in legal reasoning and statutory interpretation. Research highlights three essential requirements for legal AI:

1. Jurisdiction-specific adaptation.
2. Hybrid retrieval and reasoning mechanisms.
3. Transparent collaboration between humans and AI.

LawGPT addresses these requirements through a combination of legal document retrieval, AI-based reasoning, explainable outputs, and human escalation mechanisms.

System Architecture

LawGPT follows a five-layer architecture:

1. User Interface Layer
   • Web and mobile access.
   • Supports text and voice inputs.
   • Multilingual interaction.
2. Application and Processing Layer
   • Handles query processing, tokenization, and normalization.
   • Uses frameworks such as FastAPI or Flask.
3. Knowledge Retrieval Layer
   • Combines BM25, TF-IDF, and vector similarity search.
   • Retrieves relevant legal content from Indian laws such as the Constitution, BNS, BNSS, and IT Act.
4. Legal Reasoning and Generation Layer
   • Uses fine-tuned legal language models (e.g., Mistral or LLaMA 3).
   • Applies the IRAC framework (Issue, Rule, Application, Conclusion) for structured legal reasoning.
   • Uses reranking techniques to improve relevance.
5. Explainability, Escalation, and Security Layer
   • Generates simplified legal explanations with citations.
   • Refers users to verified lawyers for complex cases.
   • Protects user data through AES-256 encryption and JWT authentication in compliance with the Digital Personal Data Protection Act, 2023.

Key Features

• Natural-language legal query assistance.
• FIR, RTI, affidavit, and legal document generation.
• IRAC-based legal reasoning.
• Explainability dashboard with legal citations.
• Referral system for licensed legal professionals.
• Multilingual support (English, Hindi, and regional languages).
• Secure and privacy-compliant data handling.
• Continuous improvement through user feedback.

Methodology

The workflow includes:

1. Capturing user queries through text or voice.
2. Preprocessing and generating semantic embeddings.
3. Performing hybrid document retrieval using lexical and semantic search.
4. Optimizing relevance through reranking and redundancy reduction.
5. Using a legal LLM to generate legally grounded responses with citations and explanations.

Conclusion

LawGPT represents a step towards democratizing legal assistance by blending AI-driven plain-language legal advice with structured escalation to law firms. Inspired by models like ChatLaw, CHRExpert, Aalap, and Legal Assist AI, it prioritizes accessibility, fairness, and legal compliance. With careful design, LawGPT could significantly improve access to justice, particularly for underrepresented and resource-constrained populations. LawGPT represents a major step toward digital transformation in law and governance. By integrating advanced AI retrieval and reasoning models, it simplifies legal complexities into citizen-friendly insights. The system’s transparent, explainable responses make it both educational and reliable, while its architecture ensures scalability and compliance with legal standards. Future research will focus on expanding datasets to include more regional laws, integrating speech-to-text features for accessibility, and developing mobile apps for offline use. Partnerships with law schools and bar councils will further enhance data validity and ethical oversight. Ultimately, LawGPT lays the foundation for a new era of AI-driven justice—bridging citizens and institutions through equitable, explainable, and trustworthy technology.

References

[1] Cui, Y., Zhao, S., & Zhang, W., “ChatLaw: A Multi-Agent Collaborative Legal Assistant,” arXiv preprint arXiv:2306.16092, 2023. [2] Al-Shareef, A., “CHRExpert: AI-driven Court of Human Rights Assistant,” Journal of AI and Law, vol. 31, no. 2, pp. 214–230, 2023. [3] Siino, M., Spitale, M., & Ruffo, G., “Exploring Large Language Model Applications in Law: A Systematic Literature Review,” Frontiers in Artificial Intelligence, vol. 6, p. 123456, 2024. [4] Tiwari, A., Nair, R., & Gupta, K., “Aalap: An AI Legal Assistant for Indian Judiciary Using Large Language Models,” Proceedings of the IEEE Conference on Emerging Technologies in AI, pp. 188–197, 2024. [5] Gupta, R., Jha, V., & Sinha, D., “Legal Assist AI: Transformer-Based Domain-Specific Legal Reasoning Model,” International Journal of Data Science and AI Applications, vol. 12, no. 1, pp. 55–66, 2024. [6] Robertson, S., & Zaragoza, H., “The Probabilistic Relevance Framework: BM25 and Beyond,” Foundations and Trends in Information Retrieval, vol. 3, no. 4, pp. 333–389, 2009. [7] Ramos, J., “Using TF-IDF to Determine Word Relevance in Document Queries,” Proceedings of the First Text Retrieval Conference (TREC-1), 2003. [8] Li, S., Zhou, J., & Ma, Z., “BAAI/bge-reranker-large: A High-Performance Cross-Encoder for Semantic Reranking,” arXiv preprint arXiv:2310.15062, 2023. [9] Ukey, P., & Naik, A., “Legislative Gaps in India’s AI Regulation: A Legal and Ethical Perspective,” Indian Journal of Law and Technology, vol. 19, no. 1, pp. 12–25, 2024. [10] Thangaraj, S., Karthick, R., & Kumar, V., “AI-Driven Chatbot for Justice: Bridging the Legal Accessibility Gap,” IEEE Access, vol. 12, pp. 111345–111358, 2024. ?

Copyright

Copyright © 2026 Ved Jain, Adhikesh Pawar, Smruti Bhosale, Janvi Darve, Dr. S. L. Bangare. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.