# Bail is the rule, jail is the exception: Article 21 explained # Bail is the rule, jail is the exception: Article 21 explained **TL;DR:** "Bail is the rule, jail is the exception" is a constitutional norm rooted in Article 21, traced from *State of Rajasthan v. Balchand* (1977) 4 SCC 308 through *Satender Kumar Antil v. CBI* (2022) 10 SCC 51. But the rule is not absolute. Statutory embargoes like Section 43D(5) of the UAPA and Section 37 of the NDPS Act force courts to weigh the doctrine against the gravity of the charge. The 2026 Delhi-riots UAPA bail batch shows exactly how that balance is struck: as reported by LiveLaw and Bar and Bench in 2026, the alleged architects were denied bail while the alleged local facilitators were released on strict conditions. --- ## On this page - [What the doctrine actually means](#what-the-doctrine-actually-means) - [Where the rule comes from: Article 21 and the Constitution Bench line](#where-the-rule-comes-from-article-21-and-the-constitution-bench-line) - [The modern restatement: Satender Kumar Antil](#the-modern-restatement-satender-kumar-antil) - [The triple test courts apply](#the-triple-test-courts-apply) - [When statute fights the rule: UAPA section 43D(5)](#when-statute-fights-the-rule-uapa-section-43d5) - [When statute fights the rule: NDPS section 37](#when-statute-fights-the-rule-ndps-section-37) - [The 2026 Delhi-riots bail batch: the doctrine in action](#the-2026-delhi-riots-bail-batch-the-doctrine-in-action) - [Rule versus exception: a side-by-side reading](#rule-versus-exception-a-side-by-side-reading) - [What this means for your bail application](#what-this-means-for-your-bail-application) - [Frequently asked questions](#frequently-asked-questions) - [How to research this topic further](#how-to-research-this-topic-further) ## What the doctrine actually means If you remember one line from Indian bail law, it is this: bail is the rule, jail is the exception. The phrase sounds like a slogan. It is not. It is a working presumption with constitutional weight, and it controls how a judge approaches a bail plea before trial. The starting point is simple. A person who has not been convicted is presumed innocent. Pre-trial detention is not punishment. It is a precaution, justified only when the court has concrete reasons to fear that releasing the accused will defeat the trial. So the default leans towards liberty, and the prosecution carries the burden of showing why custody is necessary. That default flows from Article 21 of the Constitution, which guarantees that no person shall be deprived of life or personal liberty except according to a procedure established by law. Over five decades, the Supreme Court has read this guarantee to mean that liberty is the norm and incarceration the exception that must be justified, not the other way round. Here is the catch that trips up most readers. Parliament has carved out special offences where it has reversed the default by statute. In cases under the Unlawful Activities (Prevention) Act and the Narcotic Drugs and Psychotropic Substances Act, the law imposes extra conditions that the accused must clear before a court can even consider granting bail. The doctrine does not vanish in these cases. It collides with the statute, and the court has to reconcile the two. That collision is the real subject of this post. ## Where the rule comes from: Article 21 and the Constitution Bench line The doctrine has a clear genealogy. In *State of Rajasthan v. Balchand* (1977) 4 SCC 308, Justice V.R. Krishna Iyer crystallised the principle in words that have been quoted ever since. The Court held that "the basic rule of our criminal justice system is bail, not jail." That single sentence converted a humane instinct into a doctrine. A year later, in *Gudikanti Narasimhulu v. Public Prosecutor* (1978) 1 SCC 240, the Court returned to the theme and grounded bail in the presumption of innocence. The judgment treated the question of liberty before trial as a matter of constitutional principle rather than judicial discretion exercised in a vacuum. Detention had to answer to a standard, and that standard was Article 21. The line then ran through the white-collar context. In *Sanjay Chandra v. CBI* (2012) 1 SCC 40, arising out of a high-value economic offence, the Court held that the seriousness of the charge alone does not justify keeping an undertrial in custody for an indefinite period. The object of bail is to secure the attendance of the accused at trial, not to inflict anticipated punishment. Gravity matters, but it cannot become a permanent veto on liberty. What ties these cases together is the refusal to treat any class of offence as automatically jail-only. The Court kept insisting that the question is always specific to the accused, the evidence, and the risk, never a blanket rule that this category of person stays in. To see how courts apply the same liberty-first instinct at the pre-arrest stage, read our explainer on [anticipatory bail under BNSS Section 482](/blog/anticipatory-bail-bnss-section-482), which carries the doctrine into the world of pre-arrest protection. ## The modern restatement: Satender Kumar Antil The most important recent restatement is *Satender Kumar Antil v. CBI* (2022) 10 SCC 51. If *Balchand* coined the doctrine, *Satender Kumar Antil* operationalised it for the current era. The Court said it plainly: "Grant of bail is the rule and refusal is the exception." But the judgment did far more than repeat a famous line. It built a framework. The Court sorted offences into four categories, often labelled A, B, C and D, that determine how a bail plea should be handled depending on the nature of the offence, whether arrest was even necessary, and which statutory regime applies. Category A covers ordinary offences punishable with imprisonment up to seven years. Category B and Category C cover graver offences and those carrying death or life imprisonment. Category D deals with the special statutes, the UAPA, the NDPS Act and similar laws that carry their own bail conditions. The genius of the scheme is that it tells a judge where on the spectrum a case sits before the merits are even reached. *Satender Kumar Antil* also tied bail back to arrest discipline. Where the investigating agency has not complied with the arrest safeguards, now found in Section 35 of the Bharatiya Nagarik Suraksha Sanhita, 2023, which replaced Sections 41 and 41A of the old Code of Criminal Procedure, the accused is ordinarily entitled to be released. The message to police was blunt: arrest is not the default, and an unnecessary arrest weakens the State's case against bail. The transition from the CrPC to the new criminal codes is covered in detail in our guide to the [new criminal laws under BNS, BNSS and BSA](/blog/new-criminal-laws-bns-bnss-bsa). ## The triple test courts apply Strip away the categories and the analysis comes down to three questions. Courts call this the triple test, and it is the engine of every reasoned bail order. First, is the accused a flight risk? If there is a real chance the person will abscond and not stand trial, that weighs against bail. Roots in the community, a fixed address, family, employment and the surrender of a passport all cut the other way. Second, will the accused tamper with evidence? If documents, devices or physical evidence are still being collected, and the accused is placed to destroy or alter them, custody may be justified until that risk passes. Third, will the accused influence or intimidate witnesses? Where witnesses are vulnerable, or the accused holds power over them, the court may keep the person in until critical testimony is recorded. These three concerns, flight, tampering, and witness influence, are the legitimate grounds for refusing bail. Notice what is missing. The mere seriousness of the allegation is not on the list as a standalone ground. *Sanjay Chandra* and *Satender Kumar Antil* both warn against using gravity as a substitute for a real, particularised finding of risk. A court that refuses bail must point to the risk, not just the charge. ## When statute fights the rule: UAPA section 43D(5) Now the doctrine meets its hardest test. Section 43D(5) of the Unlawful Activities (Prevention) Act tells the court that bail shall not be granted if, on a reading of the case diary or the police report, there are reasonable grounds to believe that the accusation is prima facie true. Read that carefully. The provision flips the usual inquiry. In an ordinary case the court asks whether there is a risk that justifies custody. Under Section 43D(5), the court first asks whether the accusation looks prima facie true on the material, and if it does, the statutory bar engages and bail is ordinarily refused. The presumption of innocence does not disappear, but the threshold for release climbs steeply. This is where the constitutional doctrine and the statute pull against each other. Liberty under Article 21 says lean towards release. The UAPA embargo says do not release if the accusation reads as prima facie true. Courts have not resolved this by ignoring either side. They have held that the statutory bar is real and must be respected, while also recognising that a long, grinding period of pre-trial custody, with trial nowhere in sight, can itself violate Article 21 and justify bail despite the embargo. The longer the incarceration without progress, the heavier the constitutional thumb on the liberty side of the scale. So the UAPA analysis runs on two tracks at once. On one track, the court tests the strength of the accusation against the statutory standard. On the other, it watches the clock, because prolonged detention without trial is a constitutional harm in its own right. Both tracks feed the final decision. ## When statute fights the rule: NDPS section 37 The NDPS Act sets up a similar but distinct barrier. Section 37 imposes what courts call the twin conditions for bail in cases involving a commercial quantity of narcotics. The conditions are cumulative, not alternative. The court must be satisfied, first, that there are reasonable grounds to believe the accused is not guilty of the offence, and second, that the accused is not likely to commit any offence while on bail. Only if both are met can bail be considered, and even then in addition to the ordinary considerations. The Supreme Court confirmed the strictness of this scheme in *State of Kerala v. Rajesh* (2020) 12 SCC 122, holding that the Section 37 conditions are mandatory and that the standard is more onerous than ordinary bail. For commercial quantity, the embargo is not a formality. It is a real gate the accused must pass. The 2026 NDPS jurisprudence has reinforced this. In *State of Punjab v. Balraj Singh @ Billa*, 2026 INSC 618, decided on 2 June 2026 by a Bench of Sanjay Karol and N. Kotiswar Singh, JJ, the Supreme Court set aside a High Court order that had granted bail in a commercial-quantity heroin case, holding that the Section 37 twin conditions are mandatory and cumulative and had not been satisfied. The Court framed the stakes in striking terms: "Should there be any conflict between the sovereignty of country and personal liberty, undoubtedly, the former shall prevail, particularly, when a war is waged against the nation, be it in the form of supply of drugs, which vitally affects the national economy and health of the people." We unpack that judgment in full in our analysis of [NDPS bail and the sovereignty versus liberty balance](/blog/ndps-bail-national-sovereignty-liberty). The takeaway is that under both the UAPA and the NDPS Act, the doctrine does not give the accused an automatic right to bail. It gives the accused a starting presumption that the statute can override on a proper finding, and that the passage of time can revive. ## The 2026 Delhi-riots bail batch: the doctrine in action The clearest recent illustration of how courts balance the rule against the UAPA embargo is the Delhi-riots bail batch decided in 2026. As reported by LiveLaw and Bar and Bench in 2026, the Supreme Court heard a batch of bail pleas by several accused in the larger conspiracy case arising out of the 2020 Delhi riots, prosecuted under the UAPA, and delivered a split outcome that maps the doctrine onto facts. The lead matter in the batch included *Gulfisha Fatima v. State (NCT of Delhi)*. A neutral citation has been reported but not independently confirmed as the official number, so it is referred to here by case name, date and the reporting source rather than by an asserted citation. Treat the reported neutral number with caution and verify it against the official record before relying on it in a filing. The outcome is what matters for the doctrine. Umar Khalid and Sharjeel Imam were denied bail. Gulfisha Fatima, Meeran Haider, Shifa Ur Rehman, Mohd Saleem Khan and Shadab Ahmed were granted bail, subject to twelve strict conditions. According to the reporting, the Court drew the line on the basis of the role attributed to each accused. Those described as the alleged architects of the conspiracy stayed in. Those described as alleged local facilitators were released, on conditions designed to address the very risks the triple test targets, flight, tampering and witness influence. Read this against Section 43D(5) and the logic becomes visible. For the accused the Court saw as central to the alleged conspiracy, the material was strong enough that the statutory bar held and bail was refused. For the accused on the periphery, the combination of weaker attributed roles and the length of pre-trial custody tilted the balance back towards liberty, and the Court managed the residual risk through twelve conditions rather than continued detention. That is the doctrine doing exactly what it is meant to do: not abolishing the embargo, but applying it accused by accused, role by role. This is a snapshot of a fast-moving matter, and the position can change as the litigation continues. For a wider view of the Supreme Court's recent output, including this batch and other liberty rulings, see our [Supreme Court round-up for May 2026](/blog/supreme-court-this-month-may-2026). ## Rule versus exception: a side-by-side reading The cleanest way to hold the doctrine in your head is to separate the ordinary regime from the special-statute regime. They run on different default settings. | Question | Ordinary offence (the rule) | UAPA / NDPS commercial quantity (the exception) | | --- | --- | --- | | Starting presumption favours liberty? | ✓ Yes, bail is the default | ✗ No, a statutory gate must be cleared first | | Burden to justify custody on the State? | ✓ Yes, State must show real risk | ✗ Largely shifts; accused must clear the embargo | | Gravity of charge alone enough to refuse? | ✗ No, risk must be shown | ✗ No, but it strongly informs the prima facie test | | Triple test (flight, tampering, witnesses) decisive? | ✓ Yes, it is the core inquiry | ✓ Applied, but after the statutory threshold | | Prolonged pre-trial custody can revive bail? | ✓ Yes, delay weighs heavily | ✓ Yes, Article 21 can override the embargo over time | | Statutory embargo (s.43D(5) / s.37) applies? | ✗ No special embargo | ✓ Yes, the embargo is the controlling provision | The pattern is consistent. In the ordinary lane the doctrine operates at full strength. In the special lane the doctrine survives, but it has to fight through a statutory gate, and its strongest ally becomes the passage of time. A short custody under a strong UAPA accusation rarely yields bail. A long custody under the same accusation, with trial stalled, often does. ## What this means for your bail application If you are preparing a bail application, the doctrine is not just background reading. It dictates how you frame the plea, and the framing changes with the statute in play. In an ordinary case, lead with the presumption. Cite *Balchand* and *Satender Kumar Antil*, place your client in the correct category from the four-category scheme, and then meet the triple test head on. Show settled roots, offer to surrender the passport, and propose conditions that neutralise any tampering or witness-influence concern before the court raises it. In a UAPA or NDPS case, the strategy shifts. You cannot ignore the embargo, so meet it. Attack the prima facie strength of the accusation on the actual material, distinguish your client's attributed role from the principal accused as the Delhi-riots batch did, and, where custody has run long, put the clock at the centre of the argument, because prolonged detention without trial is the strongest constitutional lever you have against the statutory bar. The craft of building that argument, provision by provision, is set out in our [guide to drafting a bail application](/blog/drafting-bail-application-guide). One practical note on procedure. Speed is part of liberty. The Supreme Court has stressed that bail and anticipatory-bail orders must be pronounced without delay, and that an order quashing or refusing relief must state the allegations it is acting on, a discipline explained in our piece on why [FIR-quashing orders must state the allegations](/blog/fir-quashing-orders-must-state-allegations). A vague, unreasoned order is itself vulnerable on appeal. This post is general legal information, not legal advice. Bail turns on the specific facts, the exact charge and the material on record, so a real application needs case-specific advice from counsel. ## Frequently asked questions ### Is "bail is the rule, jail is the exception" written in any statute? No. It is a judicial doctrine, not a statutory provision. It is grounded in Article 21 of the Constitution and was crystallised in *State of Rajasthan v. Balchand* (1977) 4 SCC 308 and restated in *Satender Kumar Antil v. CBI* (2022) 10 SCC 51. Statutes like the UAPA and NDPS Act can and do modify how strongly it applies in particular categories of offence. ### Does the doctrine apply to UAPA cases at all? Yes, but in a limited and modified form. Section 43D(5) of the UAPA bars bail where the accusation appears prima facie true on the material. The doctrine still operates as a backdrop, and courts have held that prolonged pre-trial incarceration without trial can violate Article 21 and justify bail despite the embargo. The 2026 Delhi-riots batch shows the Court applying both the embargo and the liberty principle in the same set of cases. ### What are the twin conditions under Section 37 of the NDPS Act? For a commercial quantity, the court must be satisfied of two things together before granting bail: that there are reasonable grounds to believe the accused is not guilty, and that the accused is not likely to commit any offence while on bail. The conditions are mandatory and cumulative, confirmed in *State of Kerala v. Rajesh* (2020) 12 SCC 122 and reaffirmed in *State of Punjab v. Balraj Singh @ Billa*, 2026 INSC 618. ### What is the triple test for bail? It is the three-part risk assessment a court runs on a bail plea: whether the accused is a flight risk, whether the accused will tamper with evidence, and whether the accused will influence or intimidate witnesses. These are the legitimate grounds for refusing bail. The seriousness of the charge alone, without one of these risks, is not a sufficient reason to refuse. ### Can the gravity of the offence alone justify keeping someone in custody? Generally not, in an ordinary case. *Sanjay Chandra v. CBI* (2012) 1 SCC 40 held that the seriousness of the charge cannot, by itself, justify indefinite pre-trial detention. The court must identify an actual risk under the triple test. In special-statute cases, gravity feeds into the prima facie inquiry, but even there a long custody without trial can shift the balance back towards bail. ### What happened to the accused in the 2026 Delhi-riots bail batch? As reported by LiveLaw and Bar and Bench in 2026, the Supreme Court denied bail to Umar Khalid and Sharjeel Imam, and granted bail on twelve strict conditions to Gulfisha Fatima, Meeran Haider, Shifa Ur Rehman, Mohd Saleem Khan and Shadab Ahmed. The Court distinguished between the accused on the basis of the role attributed to each, refusing bail to those described as the alleged architects and releasing those described as alleged local facilitators. ### What is the four-category framework from Satender Kumar Antil? *Satender Kumar Antil v. CBI* (2022) 10 SCC 51 sorted offences into four categories to guide bail decisions. Broadly, Category A covers offences punishable up to seven years, Categories B and C cover graver and capital offences, and Category D covers special statutes such as the UAPA and NDPS Act with their own bail conditions. The category tells the court how to approach the plea before reaching the merits. ### Does delay in trial help an accused get bail? Yes, often decisively. Courts treat prolonged pre-trial incarceration without progress in the trial as a violation of Article 21 in its own right. Even under the UAPA and NDPS embargoes, a long period in custody with trial nowhere in sight can tilt the balance towards release. The longer the delay, the stronger the constitutional argument for bail. ### How do the new criminal laws change bail procedure? The Bharatiya Nagarik Suraksha Sanhita, 2023 has renumbered the bail and arrest provisions. The arrest safeguards once in Sections 41 and 41A of the CrPC now sit in Section 35 of the BNSS, and the bail provisions have moved to new section numbers. The underlying doctrine is unchanged, but you must cite the correct provision. See our overview of the [new criminal laws under BNS, BNSS and BSA](/blog/new-criminal-laws-bns-bnss-bsa) for the mapping. ### Are bail conditions like passport surrender always imposed? Not always, but they are common, and proposing them proactively strengthens an application. Conditions are how a court manages residual risk while still granting liberty. In the Delhi-riots batch, the accused who were released were placed under twelve strict conditions, which let the Court address the triple-test concerns without continued detention. ### Where can I read the leading bail judgments myself? The official Supreme Court judgment portal at [api.sci.gov.in](https://api.sci.gov.in) and the electronic Supreme Court Reports give you authoritative text for reported judgments. For statutory text of the UAPA and NDPS Act, [indiacode.nic.in](https://indiacode.nic.in) is the official source. Always verify a reported neutral citation against the official record before relying on it. ## How to research this topic further The bail doctrine is easy to state and hard to apply, because the answer always depends on the exact charge, the statutory regime and the stage of the case. Getting it right means reading the precedents in full, checking whether a 2026 neutral citation is the official number or only a reported one, and tracing how a statutory embargo has been applied to facts close to yours. If you want to trace this doctrine across decades of judgments, from *Balchand* to *Satender Kumar Antil* to the latest UAPA and NDPS rulings, you can [research Indian case law](/solutions/research) with Niyam, which searches across 72,000+ Indian judgments and surfaces the relevant passages with citations. Your queries stay private, never sold or used to train public models. [Start for ₹100](https://app.niyam.ai/register) or write to [hello@niyam.ai](mailto:hello@niyam.ai).