India: the Supreme Court’s zero-tolerance rule, Kerala’s policy, and the draft regulations

The Supreme Court’s rule, 2 July 2026

In Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., 2026 INSC 668, the National Company Law Tribunal had admitted an insolvency application relying on six authorities that were non-existent, wrongly cited or misquoted, and the appellate tribunal had confirmed the order without noticing. The respondent bank’s affidavit said its counsel had not cited them: the tribunal had found them through its own research. The Supreme Court (Narasimha and Aradhe, JJ.) set both orders aside.

“It is necessary for Courts to adopt a zero-tolerance mode for producing, citing or using AI-generated precedents without verification. It is a misconduct on the part of an advocate to cite such judgments without verification. Equally, it is a serious lapse if a judge relies on such a fake or hallucinated AI-generated material as precedents in support of the determination. We have no hesitation in declaring that such a decision is no decision in the eyes of the law, irrespective of whether such material had a direct or indirect bearing on the decision-making.”

• Who it binds“Zero tolerance for the Bar as well as the Bench to cite, refer to, or rely on such material.”
• No materialitySuch decisions are to be set aside “even if an iota of fake or hallucinated material enters the decision-making process”.
• The BarThe Bar Council of India is directed to constitute a committee, prescribe a guiding principle, and set the disciplinary action that will follow a violation.
• Not a banThe judgment “shall have no bearing on the rightful use of AI, but on the presentation or reliance on fake or hallucinated material as if it were a court precedent”.

Applied, 2 September 2026

In Gadiya v. Union of India, 2026 INSC 947, a customs officer had imposed a penalty of Rs 425,27,99,100 in an order that, the Court found on checking the authorities itself, relied on case law “either non-existent or [with] fake citations” and on real cases that did not lay down what the order said they did. Applying Pooja Ramesh Singh, the Court set aside the penalty and the Gujarat High Court order upholding it, without reaching the merits, and remitted the matter to a different officer of the same rank. The Court added that “assistance can never be substituted for adjudication. AI may well serve as training wheels but entrusting it with the pilot’s seat would be both imprudent and dangerous.”

Read with Faiz Enterprise in the Gujarat High Court two weeks earlier, India now has two decisions in a month in which the AI-generated authorities sat inside the decision-maker’s own order.

Kerala, 19 July 2025

The first High Court in India to issue a formally documented and binding set of guidelines restricting AI use in district and subordinate courts.

• Assistive onlyAI is defined as purely assistive software. It is categorically prohibited from being used “to arrive at any findings, reliefs, order or judgment”.
• The judge owns itThe onus for the content and integrity of the judicial order remains entirely on the judge.
• Approved toolsA dichotomy between general AI tools and “Approved AI Tools” officially screened by the High Court or the Supreme Court. No other instrument on this record maintains a whitelist.
• ConfidentialityExpress caution that widely used generative AI such as ChatGPT threatens confidentiality.

The approved-tools mechanism is the distinctive contribution. Everywhere else the duty is to verify whatever you used. Kerala restricts what you may use in the first place, which is a different and stricter design.

The Supreme Court draft

The Supreme Court of India circulated draft Regulations for Use of Artificial Intelligence in Courts, 2026. The consultation was extended from 20 June 2026 and closed on 15 July 2026. In Gadiya, on 2 September 2026, the Court itself recorded that the regulations are “yet to be finalised”. They remain a draft, and nothing on this site treats them otherwise.

Where India sits

Until July 2026, India looked like England and Wales: guidance to the judiciary, with the practitioner question left to the court’s inherent powers and the Bar Council. Pooja Ramesh Singh changed that. The duty to verify now comes from the Supreme Court itself, it binds advocates and judges alike, it carries a professional consequence for the advocate and a voiding consequence for the decision, and it applies without any showing that the unverified material affected the outcome. No rule of court prescribes a declaration, as Ireland’s HC 142 does; the Indian instrument is a judgment, and it reaches the decision rather than the filing.

Sources

• Supreme Court of India, Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd., 2026 INSC 668, judgment of 2 July 2026. Read in full from a copy bearing the Court’s neutral citation.
• Supreme Court of India, Vijay Ghanshyam Gadiya v. Union of India, 2026 INSC 947, order of 2 September 2026. Read in full from a copy bearing the Court’s neutral citation.
• Kerala High Court, Policy Regarding Use of Artificial Intelligence Tools in District Judiciary, published 19 July 2025. Not read at source.
• Supreme Court of India, draft regulations on the use of AI in courts; consultation closed 15 July 2026. Draft.
• Kerala policy and draft regulations: Indian legal commentary including SpicyIP, S.S. Rana & Co. and The Leaflet. This page carries no quotation from either of those two documents. Every quotation above is from the two Supreme Court judgments.

Related

• AI rules for lawyers, country by country
• Gadiya v. Union of India · Faiz Enterprise v. State Tax Officer
• Verification is becoming a duty. Evidence of verification is the missing layer.
• India · DPDP Act 2023
• England and Wales, the architecture India has now moved beyond

Checked 3 September 2026. The two Supreme Court judgments were read in full; the Kerala policy was not read at source; the Supreme Court regulations are a draft. Information, not legal advice, and not advice on Indian law.

Common questions

Does India have rules on AI in courts?

Yes, at two levels. The Supreme Court of India held on 2 July 2026, in Pooja Ramesh Singh v. Jammu and Kashmir Bank, that courts must adopt zero tolerance for producing, citing or using AI-generated precedents without verification, that citing such material without verification is misconduct by an advocate, and that a decision resting on it is no decision in the eyes of the law. It applied that rule on 2 September 2026 in Gadiya v. Union of India. Separately, on 19 July 2025 the Kerala High Court published its Policy Regarding Use of Artificial Intelligence Tools in District Judiciary, the first formally documented and binding set of guidelines in India restricting AI use in district and subordinate courts. The Supreme Court’s draft regulations on AI in courts remain in draft.

What does the Kerala High Court AI policy say?

AI is defined as purely assistive. It is categorically prohibited from being used to arrive at any findings, reliefs, order or judgment, and the onus for the content and integrity of a judicial order remains entirely on the judge. It also distinguishes between general AI tools and 'Approved AI Tools' screened by the High Court or the Supreme Court, and cautions that widely used generative AI such as ChatGPT threatens confidentiality.

Are the Supreme Court of India's AI regulations in force?

No. They are draft. The consultation was extended and closed on 15 July 2026. Anyone citing them as binding is wrong, and we mark them as draft everywhere they appear on this site.

Do Indian rules apply to lawyers or only to judges?

Both. The Supreme Court’s rule of 2 July 2026 binds the Bar as well as the Bench: it calls citing unverified AI-generated judgments misconduct on the part of an advocate, calls reliance on them by a judge a serious lapse, and directed the Bar Council of India to prescribe guiding principles and the disciplinary action that will follow a breach. The Kerala policy is narrower and is directed at the district judiciary, meaning judges and court staff.