NyayaSahaya: A Survey on AI-Powered Legal Assistance and Government Policy Simplification for Common People

Abstract

In developing nations, legal literacy continues to be a major challenge because the jargon employed in legal discourse and state policies serves as barriers to everyday citizens seeking legal help. This survey considers the space of AI-based legal support systems, particularly the document simplification and multilingual natural language processing approaches. We outline the existing legal AI systems deploying chatbots, contract analysis software, and policy simplification systems, their technical approaches, and limitations. The major challenges found include: data limitations within the legal space, the challenges of processing multilingual content, and accuracy of legal advice generated by AI. We introduce NyayaSahaya as a combined legal advice generator and government policy simplifier utility, aimed at non-technical users operating in multilingual environments. This survey adds to the understanding of AI systems which can through value and access democratize legal aid in developing nations who would likely require digital governance development efforts. Recent reports of legal professional uptake of AI systems indicate that in the face of legal literacy challenges, there is impetus for widespread adoption of these technologies by 2024 (increase to up to seventy-nine percent of legal professionals)

Introduction

Legal illiteracy remains a critical barrier in developing nations, where most citizens lack even a basic understanding of their rights or government policies. In India, this problem is intensified by bureaucratic jargon, high legal costs, and linguistic diversity, which together prevent people—especially in rural areas—from accessing legal information and services. Traditional legal systems are slow, costly, and inaccessible, leaving millions unaware of available government benefits or vulnerable to exploitation.

The post-pandemic era has amplified the demand for affordable, remote, and easily understandable legal assistance. Artificial Intelligence (AI) and Natural Language Processing (NLP) technologies offer promising solutions through AI-powered legal advisory systems that can provide 24/7, low-cost, multilingual support. However, these systems must ensure accuracy, cultural relevance, and ethical compliance, as incorrect advice could lead to severe consequences.

The study explores key research questions concerning how AI can process and simplify legal texts, assist non-expert users, and address socio-economic and cultural barriers. The research primarily focuses on English-based legal support systems designed for general citizens, emphasizing accessibility and fairness.

The literature review highlights significant progress in legal AI applications—from early chatbots like DoNotPay to sophisticated systems like ROSS Intelligence and LawGeex, which use NLP for document review, contract analysis, and case research. Transformer-based models such as LegalBERT and CaseLaw-BERT outperform generic models, achieving high accuracy in tasks like legal entity recognition, document summarization, and question answering. Moreover, AI-driven legal document simplification and summarization efforts have improved public comprehension of complex government policies, increasing citizen engagement in welfare programs by up to 25%.

For multilingual contexts, especially in India, cross-lingual and code-switched NLP models are helping bridge linguistic barriers, though challenges persist due to data scarcity, regional language diversity, and differences in legal terminology across jurisdictions.

Key technical challenges include limited annotated datasets, the complexity of legal language, and model hallucinations. Ethical and domain-specific challenges involve ensuring accuracy, preventing bias, safeguarding user privacy, and maintaining professional responsibility. AI systems must also consider socio-economic realities so that recommendations remain both legally sound and practically feasible.

Looking ahead, emerging Constitutional AI and agentic AI systems could enhance reliability and autonomy in legal services by automating tasks like form filing, will drafting, and business registration. Multimodal features—such as voice interfaces and document image analysis—can further expand access, especially for users with low literacy or limited digital experience.

Conclusion

The survey demonstrates considerable advancements in AI-based legal aid systems but also points to critical barriers that must be overcome if there is to be try widespread adoption or a significant positive impact on access to legal services. Current systems show great potential in areas such as legal document preparation, intelligent question answering, and legal text simplification. Yet, there are still significant restrictions in terms of reliability, multilanguage capability, and the ability to customize solutions to fit various cultures and contexts, which limit their full potential to serve different populations. Presently there are expectations that the legal AI marketplace will be valued at USD 4.9 billion by 2032 [41] . Thus, the need for affordable, reliable accessibility is more than ever complex. The proposed system, NyayaSahaya, addresses critical gaps identified in the literature review, by integrating government policy simplification with the ability to generate legal advice, created explicitly for multilingual environments and non-technical users that do not have formal legal training. The emphasis on Indian languages and actual culture in combination with the goal of legal context makes for useful next steps towards democratizing access to legal services for the developing world, as traditional legal services are traditionally excluded to most citizens. The NyayaSahaya platform provides a scalable, easy to use platform that understands the barriers of language and socioeconomic constraints, and allows millions to have the ability to confidently and comprehend the legal worlds they exist in. Key research directions should advocate for systems of evaluation of legal AI that are not limited to accuracy to value their operational effectiveness and the extent of user trust. The ability to create rich multilingual legal data sets will be vital as currently available data does not account for the linguistic legal richness of unique countries such as India. Ethical questions of bias and privacy will require continuous attention with the lens that legal AI can support underserved communities instead of perpetuating inequities. Legal AI systems could also influence digital governance, to strengthen citizen participation and legal compliance, governing bodies will continue advancing to deliver services digitally. Future directions should also consider researching the effectiveness of legal AI systems in longitudinal studies in the real world to better understand their actual impact on users legal outcomes and decision making processes. The development of acceptance metrics or benchmarks for legal AI systems to evaluate benchmarks across or between systems will improve development, performance, and reliability. There should also be collaborative effort from legal AI researchers, legal professionals, policy makers, to deploy legal AI responsibly with appropriate discretion and to promote access.

References

[1] Singh et al., ”Legal Literacy in Developing Nations: A Comprehensive Survey,” Journal of Legal Technology, vol. 15, no. 3, pp. 245–262, 2023. [2] M. Johnson, ”Automated Legal Services: Analysis of DoNotPay’s Impact on Consumer Legal Access,” AI & Law, vol. 30, no. 4, pp. 485–506, 2022. [3] P. Zhang and L. Martinez, ”ROSS Intelligence: AI-Powered Legal Research Platform,” Artificial Intelligence Review, vol. 54, no. 2, pp. 1123–1145, 2021. [4] K. Patel et al., ”Conversational AI in Legal Services: A Systematic Review,” Expert Systems with Applications, vol. 210, pp. 118–135, 2023. [5] R. Wilson and S. Chen, ”Benchmarking AI Contract Analysis: Human vs Machine Performance,” Computer Law & Security Review, vol. 39, pp. 105–118, 2020. [6] D. Kumar and A. Sharma, ”Simplifying Legal Documents: Approaches and Challenges,” Natural Language Engineering, vol. 25, no. 4, pp. 467–489, 2019. [7] L. Thompson et al., ”Neural Approaches to Legal Text Simplification,” Transactions on Computational Linguistics, vol. 10, [8] pp. 892–907, 2022. [9] J. Lee and M. Park, ”Automatic Summarization of Legal Documents: A Survey,” Information Processing & Management, vol. 60, no. 2, pp. 103–124, 2023. [10] N. Gupta et al., ”Government Policy Simplification Using Multi-Document Summarization,” Government Information Quar- terly, vol. 40, no. 1, pp. 45–62, 2023. [11] F. Rodriguez and K. Tanaka, ”Multilingual Natural Language Processing in Legal Contexts,” Computational Linguistics, vol. 48, no. 2, pp. 289–318, 2022. [12] V. Agarwal et al., ”Legal NLP for Indian Languages: Challenges and Opportunities,” ACM Transactions on Asian and Low-Resource Language Information Processing, vol. 22, no. 3, pp. 1–25, 2023. [13] S. Sharma and P. Joshi, ”Processing Code-Switched Legal Queries in Indian Contexts,” Language Resources and Evaluation, vol. 57, no. 1, pp. 167–189, 2023. [14] H. Wang et al., ”Named Entity Recognition in Legal Texts: A Comprehensive Survey,” Knowledge-Based Systems, vol. 245, [15] pp. 108–127, 2022. [16] C. Liu and D. Brown, ”Graph Neural Networks for Legal Information Extraction,” Neurocomputing, vol. 521, pp. 78–94, 2023. [17] T. Anderson et al., ”LegalQA: A Benchmark for Legal Question Answering Systems,” Proceedings of ACL, pp. 2456–2471, 2023. [18] I. Chalkidis et al., ”LegalBERT: The Muppets Straight Out of Law School,” Findings of EMNLP, pp. 2898–2904, 2021. [19] G. Miller and A. Davis, ”Evaluating Large Language Models on Legal Reasoning Tasks,” AAAI Conference on AI, pp. 15234–15241, 2023. [20] Y. Kim et al., ”T5-based Legal Document Generation and Simplification,” NAACL Proceedings, pp. 1789–1798, 2022. [21] E. Rossi and M. Fischer, ”Cross-lingual Transfer Learning for Legal Document Classification,” EACL Proceedings, pp. 567–578, 2023. [22] B. Taylor et al., ”Data Challenges in Legal AI: Privacy, Quality, and Availability,” AI Magazine, vol. 44, no. 1, pp. 78–92, 2023. [23] R. Verma and S. Chopra, ”Multilingual Challenges in Indian Legal AI Systems,” ICON Conference Proceedings, pp. 245– 256, 2022. [24] O. Nielsen and K. Jensen, ”Accuracy Requirements for AI Legal Advisory Systems,” AI & Society, vol. 38, no. 2, pp. 567–585, 2023. [25] Q. Washington et al., ”Algorithmic Bias in Legal AI: Sources and Mitigation Strategies,” FAccT Conference, pp. 123–134, 2022. [26] X. Chen and J. Williams, ”Constitutional AI for Legal Applications,” NeurIPS Workshop on Trustworthy AI, pp. 45–52, 2023. [27] Thomson Reuters, ”How AI is Transforming the Legal Profession,” Thomson Reuters Legal Blog, 2025. [28] Clio, ”AI-Powered Legal Practices Surge: Clio’s Latest Legal Trends Report,” Clio Press Release, 2024. [29] Paxton AI, ”How AI is Transforming Legal Research and Drafting: Trends and Best Practices for 2025,” Paxton AI Blog, 2025. [30] A. Quevedo et al., ”Legal Natural Language Processing from 2015-2022,” arXiv preprint, 2023. [31] LexisNexis, ”Generative AI and Legal Practice,” LexisNexis Survey, 2024. [32] S. Kumar et al., ”Natural Language Processing for Legal Documentation in Indian Languages,” ResearchGate, 2024. [33] A. Kanapala et al., ”Indian Legal NLP Benchmarks: A Survey,” arXiv preprint, 2021. [34] IndiaAI, ”Advancing Multilingual NLP for Indian Legal Systems,” IndiaAI Report, 2024. [35] G. Miller et al., ”AI-Powered Lawyering: AI Reasoning Models,” SSRN, 2025. [36] Stanford HAI, ”AI on Trial: Legal Models Hallucinate in 1 out of 6 Queries,” Stanford HAI News, 2024. [37] Thomson Reuters, ”Legal Issues with AI: Ethics, Risks, and Policy,” Thomson Reuters Blog, 2025. [38] American Bar Association, ”Big Data, Big Problems: The Legal Challenges of AI-Driven Data,” ABA Business Law Today, 2024. [39] IE University, ”The Future of AI in Law: Trends and Innovations to Watch,” IE Uncover, 2025. [40] ContractPodAi, ”Future Legal Departments in 2030: AI & Automation,” ContractPodAi News, 2025. [41] McKinsey, ”AI in the Workplace: A Report for 2025,” McKinsey Report, 2025. [42] Google Research, ”Federated Learning for Multilingual NLP,” Google Research Blog, 2024. [43] Yahoo Finance, ”Legal AI Market Size to Surpass USD 4.9 Billion by 2032,” Finance Report, 2025.

Copyright

Copyright © 2025 Abhishek Pawaskar, Tejas Kunde, Yash Sawant, Prof. Deepali Shrikhande. 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.