AI-Powered Legal Research Tools and the Unauthorized Practice of Law: Where Does Assistance End and Advice Begin?
Introduction
Every lawyer has experienced the research spiral: the citation that leads to a citation, the amendment that supersedes the provision, the judgment that distinguishes the precedent you had just found. Legal research is time-consuming, expensive, and consequential. A missed authority can mean a failed argument. A misread statute can mean professional negligence liability. It is exactly this pain that AI-powered legal research tools have targeted with increasingly sophisticated solutions.
Products like Harvey, CoCounsel, LexisNexis’s Lexis+ AI, and their Indian counterparts now offer not just retrieval but analysis: summarisation of case law, identification of precedential strength, prediction of judicial outcomes, drafting of legal arguments, and in some configurations, responses to open-ended legal questions posed by non-lawyers. The convenience is undeniable. The legal risks are substantial, and among them, the risk of enabling the unauthorised practice of law is perhaps the most structurally significant.
In India, the Advocates Act, 1961, reserves the practice of law to enrolled advocates. The unauthorised practice of law is prohibited under Section 29, and Section 45 makes it a criminal offence. What constitutes the practice of law has been interpreted broadly by Indian courts to include not merely appearing before courts but advising on legal rights, preparing legal documents, and providing opinions on legal questions. If an AI tool does any of these things, the question of whether its developer, deployer, or the non-lawyer user is engaged in unauthorised practice is legally unresolved and commercially urgent.
Legal Framework
The Advocates Act, 1961, provides that only enrolled advocates are entitled to practice before courts and tribunals and to advise clients on legal matters for a fee. The Act does not define the practice of law with the precision that might resolve the AI question, and the Bar Council of India has not issued formal guidance on AI-powered legal services.
The distinction between legal information and legal advice has been used in many jurisdictions to demarcate the boundary of unauthorised practice. Legal information, the retrieval and presentation of laws, cases, and legal materials without application to a specific individual’s circumstances, is generally treated as permissible for non-lawyers. Legal advice, the application of legal principles to a specific situation with a recommendation for a particular course of action, is reserved for qualified lawyers. This distinction collapses when an AI tool is instructed by a user to provide personalised guidance.
The Consumer Protection Act, 2019, and the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, may both be relevant. A legal AI platform that provides advice to consumers might be characterised as a service provider liable for deficiencies under the former. Under the latter, a platform that generates legal content may have safe harbour concerns if its outputs constitute professional services rather than mere information.
Judicial Developments
The Supreme Court of India has broadly construed the practice of law in multiple decisions. In Pravin C Shah v KA Mohd Ali (2001), the court held that practice of law encompasses a wide range of activities connected with the legal profession, including advice, counsel, and assistance in legal matters. The Bombay High Court in Ex-Captain Harish Uppal v Union of India (2002) similarly treated the scope of legal practice as broad enough to cover activities that go beyond courtroom appearances.
In 2024, the Bar Council of India’s Ethics Committee issued an advisory note expressing concern about AI tools being used to provide legal advice to litigants without advocate involvement, and calling on the Bar Council to consider whether platform operators who deploy such tools were in violation of the Advocates Act.
The State Bar of California’s 2024 Guidelines on AI and the Practice of Law affirm that AI tools that apply legal principles to a client’s facts and provide a recommendation constitute the practice of law, regardless of disclaimers. Several state bar associations have disciplined lawyers for submitting AI-generated briefs containing hallucinated citations.
Contemporary Issues and Analysis
The AI legal research product landscape in India is rapidly diversifying. Beyond international platforms, domestic startups offer AI tools ranging from contract analysis to precedent search to automated legal drafting. Their user bases include both lawyers, who use them to augment their practice, and non-lawyers, including small businesses, startups, and individuals, who use them to understand their legal position without engaging counsel.
The non-lawyer user case is where the unauthorised practice concern is most acute. When a small business owner asks a legal AI tool whether their employment contract violates the Fixed-Term Employment provisions under the Industrial Relations Code, 2020, and receives a detailed answer that includes a specific compliance recommendation, a legal question has been posed, the AI has applied law to specific facts, and a recommendation has been given. If this interaction were conducted by a human non-lawyer, it would be textbook unauthorised practice.
The AI platform, however, occupies a more ambiguous position. It is not a person. It cannot be enrolled as an advocate or convicted of a criminal offence. Its developer is a corporate entity that may argue it has produced a software tool rather than a legal service. Courts have consistently held that the character of an activity is determined by its substance, not its labelling.
There is also the professional responsibility dimension for lawyers using AI tools. A lawyer who delegates legal research to an AI system and files the result without independent review may be in breach of the duty of competence under the Bar Council of India Rules. An AI-generated argument that is legally incorrect, strategically harmful, or factually misleading is the lawyer’s professional responsibility, and the AI vendor’s commercial disclaimer cannot transfer that responsibility.
Comparative and International Perspective
England and Wales have approached this through the Solicitors Regulation Authority’s guidance on AI use, which requires that any AI tool delivering what amounts to legal services must be delivered through a regulated legal services provider or fall within a recognised exemption. The Legal Services Act 2007 framework explicitly covers technology-delivered legal services as a reserved legal activity when they involve advice and representation.
The ABA Model Rules’ amendments in 2024 introduced a new commentary to Rule 5.5 clarifying that supervision obligations extend to AI systems used in legal practice, and that lawyers cannot disclaim responsibility for AI outputs through contractual arrangements with vendors.
Practical and Policy Implications
For AI legal tech companies, the unauthorised practice risk can be partially mitigated through design choices. Tools that present legal information with appropriate caveats, that refuse to respond to questions framed as requests for specific advice, and that route users to qualified counsel when the question requires application to specific facts are better positioned than those that simulate lawyer-client interaction.
For law firms, the opportunity is to use AI research tools as productivity accelerants while maintaining a clear human supervision chain. Firms that develop internal AI governance policies specifying review obligations, citation verification requirements, and accountability for AI-assisted work product will be better positioned to use these tools competitively without incurring professional responsibility exposure.
Suggestions and Reforms
The Bar Council of India should urgently issue formal rules under Section 49(1)(c) of the Advocates Act on the permissible use of AI in legal practice, including standards for advocate supervision of AI research outputs, disclosure obligations to clients when AI tools were used in their matter, and restrictions on AI tools operating in contexts that constitute practice of law without advocate involvement.
A regulatory sandbox for AI legal services, modelled on the Legal Services Board’s sandbox in England, would permit supervised experimentation by Indian legal tech companies in defined parameters, allowing data collection on actual use patterns and outcomes before regulatory categories are finalised.
Conclusion
The line between legal assistance and legal advice has always been contested, long before AI arrived to complicate the question. What AI has done is redraw that line at industrial scale, making it possible for a single platform to give millions of users something that functions like legal advice, at no cost, in seconds. Whether that is a democratisation of access to justice or an unregulated displacement of professional expertise depends entirely on the quality, accuracy, and accountability of the service. Getting the regulatory calibration right matters enormously, both for public protection and for the legitimate development of a legal technology sector that can make justice more accessible without making it more dangerous.
