Chief Justice of India Surya Kant has cautioned law schools against adopting a blanket prohibition on artificial intelligence tools, emphasising that the future of legal education cannot be secured by insulating students from technological developments. Speaking in Jodhpur, the CJI called upon National Law Universities to take a leading role in shaping the relationship between law and technology and stressed that students must be trained not merely to use technological tools, but to question, verify and critically assess what those tools produce. The intervention comes at a significant moment for legal education, when generative AI is rapidly changing legal research, drafting, translation, document review and academic writing, while simultaneously raising difficult questions concerning accuracy, confidentiality, authorship, professional responsibility and the future role of human judgment.
The CJI’s observation is important because the instinctive response of educational institutions to generative AI has often been to treat it primarily as an academic-integrity problem. Law schools may be tempted to prohibit students from using AI because of concerns that assignments can be generated within seconds, citations may be fabricated, and students may submit machine-produced material without understanding the underlying legal reasoning. Those concerns are genuine. But prohibition alone does not resolve the underlying problem. A technology that is already entering courts, law firms, corporate legal departments and legal research environments cannot realistically be kept outside law schools by institutional rules. The more pressing educational question is whether law students will be taught to use the technology responsibly before they enter professional life.
The CJI’s approach, therefore, appears to move the debate away from the simplistic question of whether AI should be allowed and towards a more consequential question: what kind of lawyer should legal education produce in an age in which machines can generate legal information, arguments and drafts almost instantaneously? The answer cannot be a lawyer who blindly relies upon technology, but neither can it be a lawyer who has never been trained to understand the technology that will increasingly form part of everyday legal practice.
The distinction between using AI and relying upon AI is central to this debate. A student may legitimately use an AI system to identify broad research themes, organise information, simplify a complex statutory provision or generate preliminary questions for further research. But the resulting material cannot be treated as a legal authority merely because it appears coherent or sophisticated. Every proposition must ultimately be traced back to the statute, judgment, regulation or other primary source from which the legal position is actually derived. In other words, AI can assist in discovering the road, but the lawyer remains responsible for checking whether the road actually leads to the destination.
That principle becomes particularly important in legal research because generative AI systems can produce information with a degree of confidence that bears little relationship to its accuracy. A system may invent a case name, provide an incorrect citation, attribute an observation to the wrong judge or combine passages from different judgments into a proposition that does not exist in any reported decision. In ordinary academic writing, an inaccurate citation may affect the quality of a paper. In litigation, the consequences can be substantially more serious. An advocate who places a nonexistent authority before a court cannot escape responsibility by explaining that the citation was generated by an artificial-intelligence system.
The legal profession has already begun encountering precisely this problem. Courts in India and abroad have seen instances where lawyers have relied upon AI-generated authorities that were either inaccurate or nonexistent. This has forced courts to confront an uncomfortable reality: technological sophistication does not necessarily produce legal accuracy. A machine can generate a convincing paragraph while being entirely wrong about the law. The danger is therefore not that AI will necessarily make lawyers less intelligent; it is that it may make poorly verified legal work appear deceptively authoritative.
This is why the CJI’s emphasis on independent judgment is particularly significant. Legal education has traditionally trained students to distinguish between a primary source and commentary, between a binding precedent and a persuasive authority, and between the ratio of a judgment and observations made in passing. Generative AI does not eliminate the need for these skills. If anything, it makes them more important.
A student who asks an AI tool to identify the law governing a particular dispute must still know how to verify the answer. If the tool provides five judgments, the student must determine whether those judgments are correctly cited, whether they remain good law, whether a later larger Bench has altered the position, and whether the factual context actually corresponds with the problem being analysed. That requires precisely the kind of legal reasoning that law schools are supposed to cultivate.
The danger, therefore, is not AI itself. The real danger is uncritical dependence upon AI. This distinction also explains why a blanket prohibition could be counterproductive. Students may be prevented from using AI within the classroom while continuing to encounter it outside the classroom, perhaps without any meaningful understanding of its risks. A student who has been taught nothing about AI may eventually enter a law firm and use it to summarise confidential client documents, draft legal opinions or conduct research without appreciating the risks involved. A student who has been taught responsible AI use, by contrast, is more likely to understand that technology is an assistant rather than an independent legal decision-maker.
The issue is particularly important for law schools because the profession is already undergoing technological transformation. Digital filing, electronic case records, online hearings, machine-assisted transcription, translation tools and increasingly sophisticated legal research systems have altered the working environment of lawyers and courts. Artificial intelligence is likely to deepen that transformation. Legal education cannot prepare students for this environment by pretending that technology does not exist.
But technological integration should not be confused with technological surrender. There are aspects of legal reasoning that cannot simply be delegated to a machine. A lawyer must understand the client’s circumstances, assess credibility, appreciate the consequences of a legal strategy and decide whether a particular course of action is ethically and legally defensible. Constitutional adjudication, for example, frequently involves questions of proportionality, dignity, fairness and competing social interests. Those questions cannot be resolved merely by identifying patterns in previous cases.
A machine may identify that a particular argument has appeared frequently in earlier judgments. It cannot thereby determine whether that argument should be accepted in a new constitutional context. It may identify relevant precedents, but it cannot assume the constitutional responsibility of deciding what the law ought to mean in circumstances that have not previously been considered.
That is why the CJI’s message that legal education must lead rather than merely adapt to technology deserves attention. Law schools should not simply teach students how to operate particular AI platforms. Those platforms will change. The more durable educational objective is to teach students how to interrogate technological systems themselves.
A technologically competent lawyer should be asking: What information has the system relied upon? Can the output be independently verified? Is the underlying source current? Could the system have reproduced an existing bias? Has confidential information been exposed? Does the platform retain the data entered into it? Is the generated material sufficiently reliable for the purpose for which it is being used? Who bears responsibility if the answer turns out to be wrong?
These are not merely questions for computer scientists. They are increasingly questions for lawyers. The issue of confidentiality illustrates this particularly well. Lawyers routinely handle documents containing privileged communications, commercial strategies, personal information, medical records and financial details. Uploading such documents into an AI system without understanding its data-handling practices can create serious professional and legal consequences. A student who learns only how to obtain faster results from AI but is not taught about confidentiality may acquire a technological skill that becomes a professional liability.
The same applies to intellectual property. Generative AI can produce text, images, summaries and other forms of content by processing vast quantities of existing material. Questions concerning copyright ownership, infringement, training data and originality are still developing. Future lawyers will inevitably be called upon to advise clients about these issues. Law schools therefore need to expose students to the legal architecture surrounding AI rather than treating it merely as an academic shortcut.
The CJI’s approach also has an important implication for the design of examinations and assignments. Traditional take-home research papers may become increasingly difficult to use as a reliable measure of independent legal ability when generative AI can produce polished prose almost instantly. That does not mean such assignments should disappear. It means law schools may need to redesign assessment methods.
A student could be required to disclose the manner in which AI was used during research. Universities could distinguish between acceptable assistance, such as brainstorming or language correction, and unacceptable substitution of the student’s own reasoning. Oral examinations could be used to test whether the student actually understands the arguments contained in a written submission. Students could be asked to critique an AI-generated legal answer, identify fabricated authorities and correct errors. Drafting exercises could be conducted under supervision, while research methodology could be assessed through the student’s ability to explain how a conclusion was reached.
Such methods would arguably make legal education stronger rather than weaker. Instead of asking only whether a student can produce a sophisticated legal document, institutions could ask whether the student can identify what is wrong with a sophisticated-looking document. That is a far more realistic professional skill in an AI-enabled legal environment.
The issue also presents an opportunity to revive an aspect of legal education that has sometimes been overshadowed by the emphasis on examination performance the discipline of primary-source reading. An AI-generated summary of a Supreme Court judgment may be useful as a preliminary orientation, but it cannot substitute for reading the judgment itself. The factual matrix, submissions, statutory provisions, previous authorities and precise reasoning may all disappear in a summary.
For law students, this is especially important because the difference between a correct and incorrect understanding of a judgment often lies in a qualification contained in a few paragraphs. The ratio of a decision cannot safely be reconstructed merely from a machine-generated summary. A student who accepts such a summary without checking the original judgment risks developing an understanding of the law that is both confident and wrong.
The same concern applies to statutes. Legislation changes. Provisions are amended, substituted, repealed and brought into force on different dates. Notifications and rules may alter the practical application of statutory provisions. An AI system that gives a broadly accurate explanation of a statute may nevertheless be relying on an outdated legal position. The lawyer’s responsibility to verify the current law therefore remains indispensable.
This is where the principle of human-in-the-loop legal research becomes particularly valuable. Technology may undertake the initial work of locating, sorting or summarising information, but a human lawyer must remain responsible for verification, interpretation and final application.
The CJI’s comments also raise the larger question of what legal education itself should mean. If legal education is merely the memorisation of statutes and judgments, AI represents a serious challenge because machines can retrieve and reorganise information much faster than humans. But if legal education is understood as the development of judgment, analytical ability, ethical responsibility and constitutional understanding, technology becomes a tool rather than a competitor.
The lawyer’s value does not lie simply in knowing that Section X of a statute exists. It lies in understanding how that provision applies to a particular factual situation, how it interacts with other provisions, how courts have interpreted it and what legal and practical consequences follow from invoking it. Artificial intelligence may increasingly assist with the first stages of that process. It cannot relieve the lawyer of responsibility for the last stages.
There is also an important equality dimension to the debate. Access to technology has historically been uneven across the legal profession. Students at institutions with extensive research facilities, expensive databases and technologically sophisticated faculty may enjoy advantages over those studying in institutions with fewer resources. Properly implemented, AI could actually reduce some of these disparities by providing students with affordable assistance in organising research and understanding difficult material.
A blanket ban could have the opposite effect if technologically sophisticated students continue using AI independently while students in institutions that impose strict prohibitions remain unfamiliar with it. The result could be a new form of professional inequality between lawyers who understand emerging technologies and those who have been excluded from learning about them. The more appropriate institutional response is therefore to democratise responsible AI literacy.
Every law student should ideally graduate with an understanding of the basic capabilities and limitations of generative AI, the risks of fabricated authorities, the importance of source verification, confidentiality requirements, academic-integrity rules and the broader legal questions raised by automated systems. Such knowledge should become part of professional competence rather than an optional technological skill.
The role of faculty is equally important. Professors themselves will need to understand how AI is being used if they are expected to regulate it. An educator cannot meaningfully distinguish between legitimate research assistance and academic misconduct without understanding what these systems can actually do.
Universities may therefore need institutional AI policies that are precise rather than absolute. Instead of simply stating that “AI is prohibited”, policies could identify situations in which AI is permitted, situations in which disclosure is required, categories of information that must never be uploaded, and circumstances in which machine-generated material cannot be submitted as the student’s independent work. Such a framework would teach students something more valuable than compliance: it would teach them professional judgment.
The same principle should govern future lawyers’ use of AI in practice. A law firm may use AI to review thousands of pages of documents in a commercial dispute, but a lawyer must still determine whether the system has missed relevant material. A corporate legal department may use AI to identify unusual contractual clauses, but counsel must still assess whether the identified clause creates an unacceptable legal risk. An advocate may use AI to generate a first draft of a pleading, but the advocate remains responsible for every statement ultimately placed before the court. Professional responsibility cannot be transferred to an algorithm.
This is particularly important because courts operate on trust. Advocates are expected to present authorities accurately and assist the court rather than mislead it. If AI becomes increasingly embedded in legal research, the professional duty of verification becomes more not less important.The Supreme Court’s own technological journey provides a broader institutional context. The judiciary has increasingly adopted digital systems for filing, case management, transcription, translation and access to judgments. Artificial intelligence is naturally likely to become part of that technological ecosystem. But the legitimacy of adjudication ultimately rests on human judicial responsibility.
The same principle should govern legal education. Technology may make legal work faster; it should not make legal reasoning disposable. There is also a constitutional dimension to the issue. Lawyers are not merely service providers. They are participants in the administration of justice and, in many cases, defenders of constitutional rights. Future advocates will increasingly confront questions concerning algorithmic discrimination, privacy, surveillance, automated decision-making, deepfakes, intellectual property and liability for AI-generated harm.
A law school that excludes AI entirely from its educational environment risks producing graduates who are unfamiliar with some of the most significant legal questions they will encounter. The CJI’s intervention therefore represents a broader call for legal education to move from technological resistance to technological literacy.
But technological literacy should be accompanied by technological scepticism. Students must be taught that an AI system is not an oracle. Its fluency is not proof of accuracy. Its confidence is not evidence. Its ability to produce a citation does not establish that the cited case exists. Its ability to summarise a judgment does not mean that it has understood the ratio. And its ability to generate a legal argument does not mean that the argument is legally sustainable. These distinctions should become part of basic legal training.
Perhaps the most useful exercise for a modern law school would therefore be to ask students not simply to use AI but to cross-examine it. Give students an AI-generated legal opinion containing several subtle errors and require them to identify each one. Ask them to verify every authority, locate the original judgments, identify outdated provisions and explain why the machine’s reasoning fails. Such an exercise would combine technology with traditional legal methodology. It would also reinforce a fundamental truth about the profession: lawyers are not rewarded merely for producing answers. They are expected to determine whether the answer can withstand scrutiny.
The debate over AI in law schools is consequently not really a debate about technology. It is a debate about educational philosophy. Should law schools train students to reproduce information, or should they train them to interrogate information? If the latter is the objective, AI does not eliminate the need for legal education. It makes good legal education even more important.
The CJI’s remarks should therefore not be interpreted as an invitation to permit unrestricted AI use. They represent a more nuanced position: do not prohibit what students need to understand; teach them how to use it responsibly and how to recognise when it cannot be trusted.
That distinction will become increasingly important as AI systems become more capable. The more convincing the output becomes, the more difficult it may be for an inexperienced user to recognise an error. Future legal professionals will therefore need not only technological competence but a deeply ingrained habit of scepticism and verification.
There is an additional ethical dimension. Law students are eventually entrusted with clients whose lives, liberty, property and reputation may depend upon legal advice. If AI is used merely to save time, the temptation will always be to accept its output without sufficient scrutiny. Legal education must instead instil the principle that efficiency can never override accuracy where legal rights are at stake.
A lawyer who spends ten minutes verifying an AI-generated research result may be more professionally competent than one who produces an apparently perfect opinion in thirty seconds without checking it.
The real measure of technological advancement in the legal profession should therefore not be how quickly lawyers can produce documents. It should be whether technology enables lawyers to devote more time to the parts of legal work that require distinctly human abilities strategy, judgment, advocacy, client counselling, negotiation and ethical decision-making.
That is the opportunity that AI presents to legal education. If used properly, artificial intelligence could reduce the mechanical burden of legal research and allow students to concentrate more deeply on analysis. It could help students navigate large bodies of case law, compare legal positions and identify questions that require further investigation. But if used carelessly, it could produce a generation of lawyers who are extraordinarily efficient at generating legal language while becoming progressively weaker at understanding the law behind that language.
The choice will depend largely on educational institutions. Law schools should not respond to AI by pretending it does not exist. Nor should they embrace it uncritically because it promises efficiency. They should incorporate it into legal education with clear ethical boundaries, strong verification requirements and an unwavering emphasis on independent reasoning.
The CJI’s message ultimately points towards a simple but consequential principle: the future lawyer must be technologically capable without becoming intellectually dependent on technology. Artificial intelligence can search faster than a student, summarise faster than a student and draft faster than a student. What it cannot legitimately do is assume the professional responsibility of the lawyer. The law student must still read, question, verify, reason and decide.
That is why the most appropriate response to AI in legal education is neither fear nor blind enthusiasm. It is disciplined engagement. Law schools should teach students what AI can do, what it cannot do, where it fails, what legal risks it creates and how its output must be tested against primary legal sources.
The future of legal education, in that sense, should not be about producing lawyers who compete with machines at retrieving information. It should be about producing lawyers who know how to use machines without surrendering their judgment to them.
The CJI’s intervention is therefore timely. The legal profession is entering a technological phase in which the most valuable skill may no longer be the ability to find an answer quickly, but the ability to determine whether the answer deserves to be believed. For law schools, that should become a central educational objective. AI may increasingly assist the lawyer, but the final responsibility for the law, the argument and the consequences of relying upon it must remain with the human professional.
