{"slug": "india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making", "title": "India Supreme Court Proposes Ban on AI in Judicial Decision-Making", "summary": "India's Supreme Court has proposed draft regulations, released for public consultation in June 2026, that would ban AI tools from judicial decision-making while permitting them only for assistive tasks, following a July 2, 2026 ruling in Pooja Ramesh Singh vs Jammu and Kashmir Bank Ltd. and Anr. that set aside a National Company Law Tribunal judgment and called AI-generated errors an \"invisible,\" \"insidious,\" and \"catastrophic\" contamination of the justice delivery system. The draft rests on five principles — human primacy, transparency, accountability, data protection, and judicial autonomy — as India's courts carry a backlog of more than 60 million pending cases, and a 2024 Stanford and Yale study found legal-specific AI tools hallucinate at rates of 17 per cent to 34 per cent, with a global tracker recording more than 1,725 court penalties and fines reaching 110,000 US dollars in a single case as of mid-2026.", "body_md": "**October 10, 2026, (Inside AI)** — India's Supreme Court has drawn a hard line on artificial intelligence in the courtroom, proposing a ban on AI tools in judicial decision-making while allowing them only for assistive tasks. The draft regulations, released for public consultation in June 2026, rest on five principles: human primacy, transparency, accountability, data protection, and judicial autonomy. The move follows a scathing July 2, 2026 ruling in which the court set aside a National Company Law Tribunal judgment in *Pooja Ramesh Singh vs Jammu and Kashmir Bank Ltd. and Anr.*, describing AI-generated errors as an \"invisible,\" \"insidious,\" and \"catastrophic\" contamination of the justice delivery system.\n\nThe stakes are enormous. India's courts carry a backlog of more than **60 million** pending cases. AI promises relief through faster research, drafting, case management, transcription, and risk flagging. Yet the same technology has already produced documented failures worldwide, from biased risk scores to fabricated legal citations. The question is no longer whether AI belongs in law. It is where its role ends and who answers when it fails.\n\n## Foreign Courts Show Both Promise And Peril\n\nChina's Hangzhou internet courts report that AI-assisted case management has cut hearing times by more than **50 per cent**. Estonia uses an automated system called **Salme** to handle transcription and redaction, freeing judges for decision-making. In India, the Sikkim High Court became the first fully paperless court, and district courts are testing AI for assistive functions.\n\nThe cautionary tales are equally concrete. A [risk assessment tool used in the United States](https://www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing) assigned harsher scores to Black defendants. A Dutch welfare-fraud detection system penalized applicants based on nationality. Such flawed predictions can compound into heavier policing, denied rights, and breached privacy.\n\nHallucinations worsen the problem. A language model does not retrieve case law the way a legal database does. It generates the statistically probable next word. Because legal citations follow a predictable format, a fabricated one can look entirely authentic. A **2024** study by **Stanford** and **Yale** found that even legal-specific AI tools carry hallucination margins of **17 per cent to 34 per cent**.\n\nCourts in the **United States**, **United Kingdom**, **Canada**, and the **European Union** have responded with sanctions. A global tracker recorded more than **1,725** penalties as of mid-2026, with fines reaching **110,000 US dollars** in a single case. India's Supreme Court has demanded a \"zero-tolerance\" approach to bad outputs.\n\n## Accountability Must Stay Human\n\nThe draft regulations propose prohibiting AI in judicial decision-making and [limiting it to assistive functions](https://insideai.news/news/ai-policy-and-regulation/samarkand-declaration-ai-judges/13655/). But enforcement remains unclear. Recent instances of foundational AI models bypassing containment constraints in weak sandboxes have raised alarms about rogue behavior. The authors of the source analysis argue that AI must not be unfettered. It must face safeguards around credibility, reliability, and fit-for-use evaluations.\n\nEvery citation warrants a human check against the primary source. Every AI-assisted output warrants scrutiny proportionate to what is at stake. Institutions introducing these tools owe the public a clear account of what the machine decided, what a person decided, and why the two were kept separate.\n\nThe authors call for a nimble mechanism to make and regularly review informed choices about AI use, accounting for the stakes, the nature of the dispute, and whether the use moves from assistive to decision-making territory. Legal professionals should adopt higher thresholds of diligence and duty of care. Law societies, associations, and law schools should pursue dedicated upskilling. Professional codes should be upgraded with permissible AI use cases, accountability matrices, and consequences for over-reliance.\n\nThe analysis closes with a 50-year-old observation from **Krishna Iyer J**, which the authors say remains apt: \"...sociology-cultural changes are the sources of the new values, and sloughing off old legal thought is part of the process of the new equity-loaded legality... the rule of law enshrined in our Constitution must and does reckon with the roaring current of change which shifts our social values and shrivels our deferral roots, invaded our lives and fashions our destiny.\"\n\n**Read:** **Arizona Appeals Court Tosses Sentence After AI Video of Victim Played in Court**\n\nAs courts worldwide move from reprimands to disbarment and revised conduct codes, India's consultation period will test whether its framework can keep pace with the technology it seeks to govern.", "url": "https://wpnews.pro/news/india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making", "canonical_source": "https://insideai.news/news/ai-policy-and-regulation/ai-in-judicial-decision-making/14011/", "published_at": "2026-10-10 05:07:04+00:00", "updated_at": "2026-10-10 05:28:46.292465+00:00", "lang": "en", "topics": ["ai-policy", "artificial-intelligence", "ai-ethics", "large-language-models"], "entities": ["India Supreme Court", "National Company Law Tribunal", "Pooja Ramesh Singh vs Jammu and Kashmir Bank Ltd. and Anr.", "Hangzhou internet courts", "Salme", "Sikkim High Court", "Stanford", "Yale"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making", "markdown": "https://wpnews.pro/news/india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making.md", "text": "https://wpnews.pro/news/india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making.txt", "jsonld": "https://wpnews.pro/news/india-supreme-court-proposes-ban-on-ai-in-judicial-decision-making.jsonld"}}