S VimalaArtificial Intelligence may possess intelligence, but only law can ensure responsibility, fairness, and humanity. The sentence captures one of the greatest legal questions of our age. AI is no longer a futuristic imagination. It writes, predicts, recommends, translates, diagnoses, grades, monitors, and increasingly influences decisions affecting human lives. It has quietly entered classrooms, hospitals, banks, workplaces, police systems and legal research.Yet India — a constitutional democracy founded upon justice, liberty, equality, and dignity— has no dedicated law governing AI.That legislative silence is unwise.AI excels in speed and efficiency. But it does not possess conscience. It does not understand suffering, fairness, mercy, or morality. It predicts probability, not justice.In a constitutional democracy, every exercise of power must be answerable. A judge can be questioned in appeal. A govt officer may be challenged before a court. A tribunal must give reasons for its conclusions. Administrative decisions must comply with principles of natural justice.But a silent machine cannot enter a witness box. It cannot explain itself. It cannot bear responsibility.That is why AI may assist decision-making, but it must never become the unquestionable decision-maker in matters affecting liberty, livelihood or justice.Imagine an AI system predicting whether an accused person may reoffend. If courts blindly rely upon that score for bail, liberty risks becoming a mathematical calculation. Worse still, if historical data contains bias against certain communities, technology may silently reproduce old prejudices under the appearance of scientific neutrality.This concern is not hypothetical. In the United States, courts used an algorithmic tool called COMPAS to assess criminal risk. Critics argued that defendants could not challenge how the system reached its conclusions because the software functioned like a “black box”. AI may assist judicial thinking, but it cannot replace judicial accountability.India’s present legal framework belongs to an earlier technological era. The Information Technology Act, 2000, was enacted long before generative AI, deepfakes, algorithmic decision-making, predictive analytics, or AI-generated legal and academic content became realities.The Digital Personal Data Protection Act addresses privacy concerns, but AI raises much deeper legal questions. For instance, can employers silently use AI to reject job applicants or can students use AI in research, and where does assistance become dishonesty? Can banks deny loans based on automated profiling? Can welfare schemes rely upon AI to approve or reject benefits? Can courts use AI-assisted legal research without safeguards?Without law, institutions invent their own standards. One university may reward responsible AI use; another may punish it as plagiarism. Such inconsistency breeds arbitrariness, and arbitrariness is the enemy of constitutional governance.Law exists precisely to replace uncertainty with fairness.India need not blindly imitate other nations. But it can certainly learn from them.The most balanced model comes from the European Union, whose AI Act regulates AI according to risk rather than treating every technology alike. A grammar-checking tool poses little danger. But AI used in criminal justice, recruitment, healthcare, education, or public welfare may deeply affect human lives.While EU adopts a strict risk-based model, Canada insists on explainable govt decisions. Singapore builds trust through governance frameworks and fairness-testing tools. The United States follows sectoral regulation through existing agencies. China tightly controls recommendation systems and digital content.Yet despite these differences, one shared lesson emerges: No mature legal system permits the gravest decisions concerning liberty, livelihood, or access to justice to rest entirely in a machine no one can question.This principle deserves constitutional recognition in India.AI undoubtedly offers enormous promise for Indian courts. It can summarise judgments, simplify legal research, translate documents, reduce delays, and improve access to justice.But it also poses risks. AI systems may hallucinate false precedents, distort legal reasoning, or quietly reproduce hidden bias.Therefore, courts must adopt safeguards: mandatory verification of AI-generated material, disclosure of AI assistance, confidentiality protections, and above all, human reasoning in every judicial determination.India urgently needs a dedicated Artificial Intelligence Regulation Act. Such legislation should establish an AI Regulatory Authority, require registration and audit of high-risk systems, impose transparency duties, create liability for harmful deployment, and guarantee remedies for citizens adversely affected by automated decisions.India’s constitutional promise is not merely efficient governance; it is just governance. It is not merely rapid decisions; it is fair decisions. Before machines begin shaping the future of citizens, Parliament must shape the rules for machines.A practical Indian law could classify AI systems into four categoriesProhibited AI systems: Social scoring of citizens, manipulative systems targeting vulnerable persons, unlawful mass surveillance, or automated denial of rights without human review are not permittedHigh-risk AI systems: AI used in courts, policing, healthcare, education, recruitment, banking, insurance, welfare delivery, and public administration must undergo scrutinyLimited-risk systems: Chatbots, recommendation systems and AI-generated content should carry transparency obligations so citizens know they are interacting with AIMinimal-risk systems: Low-impact tools such as spam filters may remain unregulatedSupreme Court draws a lineOn July 2, Supreme Court observed that the use of AI-generated “hallucinated” judicial precedents is “catastrophic” to the judicial process, while setting aside a National Company Law Tribunal (NCLT) order after finding that it had relied on fictitious AI-generated case laws. The bench observed courts must adopt a “zero-tolerance” approach towards AI-generated precedents without independent verification.(The writer is a former judge of Madras high court)
