Jaswinder Singh @ Jassi v. State of Punjab & Anr. (CRM-M-22496-2022)

Jaswinder Singh, also called Jassi, was in custody in connection with FIR No. 0120 of 2 June 2020 at Police Station Shimlapuri, Ludhiana. The FIR was registered for attempt to murder and related offences, and Section 302 of the Indian Penal Code was added after one of the victims died.¹ He sought regular bail from the Punjab and Haryana High Court under Section 439 of the Code of Criminal Procedure in petition CRM-M-22496-2022. The petition was heard by Justice Anoop Chitkara, who reserved judgment on 21 February 2023 and pronounced it on 27 March 2023.¹

According to the State, the complainant and the deceased, both unarmed, were waylaid by the petitioner and his accomplices. Nine of the ten injuries on the complainant came from a sharp weapon and included the amputation of a finger, and the second victim died of multiple sharp-weapon injuries. The petitioner was named in the FIR, and after arrest the police recovered a "daat" (a bladed weapon) from him during interrogation.¹ The State also relied on his criminal history: the petitioner declared two earlier cases, and the State's status report listed three FIRs, among them two attempt-to-murder cases.¹

The court first applied conventional bail considerations. It held that the petition did not show that a recidivist would not offend or abscond if released, that the role attributed to the petitioner was specific so that parity with co-accused did not help him, and that the trial was already at an advanced stage with five witnesses examined.¹

The distinctive part of the reasoning is the proposition that cruelty changes the bail calculus. Justice Chitkara wrote that "causing death itself is cruel but if the cruelty causes death, the tables turn", that "when the offense is heinous and the crime is brutal, cruelty becomes one of the factors in granting or refusing bail", and that once a court forms a prima facie view that the accused acted with cruelty, bail should ordinarily be refused, or, if granted, only after recording reasons. He drew on the U.S. Supreme Court's description of cruelty in In re Kemmler and on a criminal law textbook, and concluded that the allegations, injuries and evidence showed "ferociousness and total lack of mercy".¹²

After the operative reasoning, the order has a section headed "Post-reasoning". It states that, to "further assess the worldwide view on bail when the assault was laced with cruelty", an AI platform "trained with multitudinous data" was used, and that the following question was put to ChatGPT: "What is the jurisprudence on bail when the assailants assaulted with cruelty?"¹

The answer, which the judge reproduced in full with a footnote to the chat interface, was generic. It said the position would depend on the facts and the law of the jurisdiction, that defendants charged with violent crimes involving cruelty may be seen as "a danger to the community and a flight risk", that judges consider the severity of the assault, criminal history and strength of evidence, and that the presumption of innocence still means defendants are entitled to bail unless there are compelling reasons to deny it.¹

Paragraph 11 then limits the passage: "Any reference to ChatGPT and any observation made hereinabove is neither an expression of opinion on the merits of the case nor shall the trial Court advert to these comments. This reference is only intended to present a broader picture on bail jurisprudence, where cruelty is a factor."¹ Bar & Bench reported that the judge had used the tool "to get a broader picture of bail jurisprudence" and that the court clarified the reference was not an expression on the merits.² The petition was dismissed, and the trial court was asked to take up the case on priority and try to conclude it by 31 July 2023, on the footing that the petitioner would not seek adjournments.¹

The order does not decide any question of law about AI, and the sources reviewed show no challenge to the ChatGPT passage. Its importance is as an early and visible instance. It showed that judges would consult general-purpose chatbots, and that they might publish the result in a reasoned order without a rule on disclosure, verification or confidentiality.

Several features later became the subject of policy. The passage gives no model version or date of the query and records only a single question, so the answer cannot easily be reproduced or audited. The answer cited no authority, so a reader could not check it. And the interface used was a public cloud service, a risk Kerala's 2025 policy addresses by barring cloud-based chatbots except for approved tools and requiring an audit record of each use.⁷ An SCC Online article uses this order to illustrate how generative AI trained mostly on higher-court bail rulings could shape what a judge sees.⁴

Five months later, Justice Prathiba M. Singh of the Delhi High Court wrote in Christian Louboutin SAS v. The Shoe Boutique that ChatGPT "cannot be the basis of adjudication of legal or factual issues in a court of law" and that AI "cannot substitute either the human intelligence or the humane element in the adjudicatory process".³ Later policies in effect follow the Louboutin line: AI may help with preliminary understanding or research, while the decision stays with the judge.

The Chitkara order was not the last time Indian judges used ChatGPT in reasons. On 16 August 2024 Justice Biswajit Palit of the Tripura High Court wrote that "assistance of ChatGPT is taken" on whether a plaintiff seeking an injunction can also get damages, quoted the answer, and remanded the case for evidence on damages without treating the answer as authority.⁵ On 10 December 2024 Justice Deepak Gupta of the Punjab and Haryana High Court put a question about the accuracy of differential GPS to ChatGPT in a land-demarcation dispute and accepted the description of centimetre-level accuracy.⁶ In February 2025 a Kerala High Court Division Bench recorded that ChatGPT "provided us with the prevalent policies" on brain death.¹⁰

Institutional responses followed. Kerala issued its district-judiciary policy in July 2025,⁷ and on 5 December 2025 a Supreme Court bench led by the Chief Justice of India declined to issue directions on a plea to regulate AI in the judiciary, saying that judges were being extremely cautious and that AI would never be allowed to overpower judicial decision-making.¹¹ In April 2026 the Punjab and Haryana High Court itself directed district judges in Punjab, Haryana and Chandigarh not to use ChatGPT, Gemini, Copilot or similar tools for writing judgments or legal research, warning that any violation would be viewed seriously.⁸ On 2 July 2026 the Supreme Court declared that AI must remain under "a human in the loop at every stage" while it set aside a tribunal ruling that relied on hallucinated precedents.⁹

As of 15 October 2026 the bail order stands as a decided matter. The sources reviewed did not show any appeal or challenge directed at the ChatGPT passage, and the outcome of the trial in FIR No. 0120 of 2020 was not established for this entry. The case remains significant as among the earliest documented uses of a generative-AI chatbot's output in an Indian judicial order, and it features in commentary, such as SCC Online's April 2026 article, on why India's courts have moved from ad hoc experimentation to written rules.