Introduction
In 1977, a two-judge bench of the Supreme Court compressed the law of bail into three words: the basic rule is bail, not jail.1 The phrase has since been repeated in hundreds of judgments, and it is now so familiar that it can sound like a description of how the system actually works. It is not. It is a statement of what the system is supposed to do.
The numbers are hard to reconcile with the slogan. According to the National Crime Records Bureau, under-trial prisoners made up 72.6 per cent of India’s prison population at the end of 2024, about 3.71 lakh persons.2 The share had peaked at 77.1 per cent in 2021 and stood at about two-thirds in 2005.3 In other words, the proportion of unconvicted persons in Indian jails has stayed high through two decades of judicial exhortation. India also ranks among the countries with the highest proportion of pre-trial detainees in the world.4
This is not only a statistical concern. Every under-trial prisoner is, in law, innocent. Detention before conviction takes away a person’s livelihood, separates families, exposes the detainee to overcrowded and often violent institutions, and weakens the ability to prepare a defence. When the detention outlasts the eventual sentence, or ends in acquittal, the State has punished a person it could not prove guilty. The question this article asks is why a legal order that so firmly prefers liberty produces so much custody. The answer, it suggests, lies less in the words of the law than in the institutions and incentives through which those words pass.
The constitutional foundations of pre-trial liberty
The constitutional starting point is Article 21, which forbids the deprivation of personal liberty except according to procedure established by law.5 After Maneka Gandhi v. Union of India, that procedure must be fair, just and reasonable, and not arbitrary, fanciful or oppressive.6 A procedure that keeps a person in custody for years without trial cannot easily meet that standard, whatever the statute says.
Article 22 adds specific guarantees at the moment of arrest. A person arrested must be told the grounds of arrest, must be allowed to consult and be defended by a lawyer of choice, and must be produced before a magistrate within twenty-four hours.7 Article 39A, though a Directive Principle, requires the State to ensure that opportunities for securing justice are not denied to any citizen by reason of economic or other disabilities, and to provide free legal aid.8 Read together, these provisions describe a person who is arrested, brought quickly before a judge, represented and, unless there is a good reason, released.
The Court gave these guarantees practical content in D.K. Basu v. State of W.B., where it laid down requirements for every arrest: a memo of arrest attested by a witness, information to a relative or friend, medical examination, and an entry in a diary at the place of detention.9 Many of these requirements were later written into the Code. Their purpose was to make arrest visible and therefore reviewable. A person whose arrest is recorded, whose family knows where he is and who is examined by a doctor is in a far better position to seek bail than one who simply disappears into a police lock-up. The constitutional design, in short, assumes that the moment of arrest is also the moment when the question of release begins.
The Supreme Court built the modern law of bail on this foundation in a remarkable series of judgments in the late 1970s, most of them written by Justice V.R. Krishna Iyer. In Gudikanti Narasimhulu v. Public Prosecutor, he described bail as a question touching personal liberty and therefore one that engaged Article 21, and warned that refusing bail is a serious matter because it is a denial of freedom.10 In Moti Ram v. State of M.P., the Court held that a magistrate should not demand sureties from a poor accused that he cannot furnish, should not insist that sureties come from the same district, and may release a person on a personal bond without sureties.11 The Court saw clearly that money bail, applied without regard to means, punishes poverty.
Later decisions confirmed that bail is not a form of punishment. In Sanjay Chandra v. CBI, the Court held that the object of bail is to secure the appearance of the accused at trial, and that it is neither punitive nor preventive.12 The gravity of the accusation is relevant, but it cannot by itself justify indefinite detention. India’s obligations under the International Covenant on Civil and Political Rights point the same way: Article 9(3) states that it shall not be the general rule that persons awaiting trial are detained in custody.13
The statutory scheme: from the CrPC to the BNSS
The Code of Criminal Procedure, 1973 divided offences into bailable and non-bailable. For bailable offences, bail was a right, and the police or the court had to release the accused on furnishing security.14 For non-bailable offences, bail was a matter of discretion for the magistrate and, more broadly, for the Sessions Court and the High Court.15 The Bharatiya Nagarik Suraksha Sanhita, 2023, which replaced the Code from 1 July 2024, keeps this architecture almost entirely.
Three statutory devices deserve particular attention because they were designed to limit detention regardless of judicial discretion. The first is default bail. If the investigation is not completed within sixty or ninety days, depending on the gravity of the offence, the accused becomes entitled to release on furnishing bail.16 The Supreme Court has described this right as an indefeasible part of Article 21 that cannot be defeated by filing a charge-sheet after the application is made.17
The second is the ceiling on detention. Section 436A of the Code, inserted in 2005, provided that an under-trial who had spent half of the maximum sentence for the offence in custody was to be released on a personal bond.18 Section 479 of the BNSS goes further for first-time offenders, who are to be released after serving one-third of the maximum sentence, and it places a duty on the jail superintendent to apply to the court.19 In August 2024, on the Union Government’s concession, the Supreme Court directed that section 479 be applied to under-trials whose cases were pending when the BNSS came into force.20
The third is the restraint on arrest itself. In Arnesh Kumar v. State of Bihar, the Court held that for offences punishable with imprisonment up to seven years, arrest must be justified against the conditions in section 41 and that a notice of appearance should ordinarily be issued instead.21 These conditions now appear in section 35 of the BNSS.22
The BNSS is not purely liberalising, however. Section 187(2) allows the fifteen days of police custody to be taken in whole or in parts during the first forty or sixty days of detention, rather than only at the start.23 Critics, among them PRS Legislative Research, have pointed out that this gives the police a reason to oppose bail during the whole of that period, since release would end the possibility of further police custody.24 A change of this kind may look technical, but in the hands of a cautious magistrate it can quietly extend detention.
The numbers behind the principle
Official data show who is detained and for how long. At the end of 2024, about a third of under-trial prisoners had been in custody for up to three months, but roughly three in ten had been held for more than a year, and 2.4 per cent for five years or more.25 Nearly half of all under-trials were between eighteen and thirty years of age.26 The data also show a consistent over-representation of persons from Scheduled Castes, Scheduled Tribes and minority communities, and of persons with little formal education, when compared with their share of the population.27
Perhaps the most telling figure is a different one. In 2022, more than fifteen lakh under-trial prisoners were released during the year, about ninety-five per cent of them on bail, while only 537 were released under section 436A.28 Bail, in other words, is granted in very large numbers. The problem is that it is granted late, and that it is often granted on terms that a significant number of prisoners cannot meet. The near-total disuse of section 436A also shows how a mandatory statutory right can wither when no one is made responsible for invoking it.
These figures also need to be read alongside the condition of the institutions in which under-trials are held. Indian prisons have for years operated above their sanctioned capacity, and the burden of overcrowding falls mainly on district jails, where under-trials are concentrated.29 Overcrowding affects access to lawyers, to family visits and to medical care, and it makes the separation of first-time detainees from habitual offenders difficult. An under-trial who enters such a prison for a minor offence may leave it with new contacts, new debts and a weakened position in the community. Detention that the law treats as a neutral holding measure is, in practice, a punishment of its own.
Why the rule fails in practice
The first reason is money. Bail in India is still largely secured by bonds with sureties. An order granting bail is of little use to a daily-wage worker who cannot find a local surety with property or produce the amount demanded. The Law Commission of India, in its 268th Report, noted that the reliance on monetary bail discriminates against the poor and recommended a greater use of personal bonds and non-monetary conditions.30 The Commission had already identified the congestion of under-trial prisoners as a problem in 1979.31 The Union Government’s scheme to support poor prisoners who cannot afford bail or fines, announced in 2023, is an acknowledgement of the same difficulty, but it depends on district committees and administrative follow-up.32
The second reason is the quality of representation at the earliest stage. Article 22 and Article 39A promise legal assistance, but an accused person produced for the first time before a busy magistrate frequently meets a legal aid lawyer for a few minutes, if at all. The first remand order often sets the course of the case. The Legal Aid Defense Counsel System introduced by the National Legal Services Authority in 2022 is an important step towards full-time public defenders, but its coverage and capacity are still uneven across districts.33
The third reason is the culture of the trial court. A magistrate or Sessions Judge who grants bail in a serious case risks criticism, while a refusal is rarely questioned. The safer course, from the point of view of the individual judge, is to leave the decision to a higher court. The Supreme Court has itself remarked on this tendency and on the resulting flood of bail matters reaching the High Courts and the Supreme Court.34 The cost of this caution is borne by the accused, who spends the interval in jail.
The fourth reason is delay in trial. A rule that favours bail is easier to apply when trials are quick. When trials take years, courts confront the choice between releasing a person accused of a serious offence for a long period or detaining a presumptively innocent person for that period. Hussainara Khatoon v. Home Secretary, State of Bihar exposed in 1979 that some under-trials had spent longer in custody than the maximum sentence for their alleged offences, and recognised speedy trial as part of Article 21.35 More than forty years later, the same pattern can still be found.
The fifth reason is administrative. Under-trials depend on the State to bring them to court. Shortages of police escorts, transfers between jails and the irregular functioning of video links mean that prisoners are frequently not produced on the date fixed, and each missed date pushes the case further away.36 Remand is often extended in a few seconds, without the prisoner being heard and sometimes without the prisoner being present at all.37 Where a prisoner has been granted bail, the verification of sureties by the police may itself take weeks. None of these delays appears in the text of the law, yet together they explain why an order of release does not always mean release.38
Special statutes and the reversed presumption
The general rule has been displaced in some of the most consequential areas of criminal law. Section 43D(5) of the Unlawful Activities (Prevention) Act, 1967 bars bail if the court finds reasonable grounds for believing that the accusation is prima facie true.39 In National Investigation Agency v. Zahoor Ahmad Shah Watali, the Supreme Court held that at the bail stage the court must accept the prosecution material broadly at face value and may not examine it in depth.40 The effect is to make bail very difficult to obtain under the statute.
Section 45 of the Prevention of Money-Laundering Act, 2002 imposes twin conditions: the court must be satisfied that there are reasonable grounds for believing that the accused is not guilty and is not likely to commit an offence while on bail.41 The Supreme Court struck down an earlier version of these conditions in Nikesh Tarachand Shah v. Union of India,42 but after Parliament amended the provision, it upheld them in Vijay Madanlal Choudhary v. Union of India.43 Similar restrictions appear in section 37 of the Narcotic Drugs and Psychotropic Substances Act, 1985.44
The NDPS Act shows how the reversed presumption reaches far beyond terrorism and high finance. Where the quantity of the drug alleged is commercial, the twin conditions in section 37 apply, and the accused must satisfy the court that there are reasonable grounds for believing that he is not guilty before trial has even begun. Many such accused are couriers or small sellers rather than organisers of the trade. For them, the classification of the quantity, which turns on a laboratory report and a seizure memo, effectively decides whether they will spend years in custody. A rule designed for the most dangerous offenders thus becomes the ordinary experience of the least powerful participants in the offence.
The constitutional courts have developed one important counterweight. In Union of India v. K.A. Najeeb, the Supreme Court held that statutory restrictions on bail do not oust the power of constitutional courts to grant bail on the ground of violation of Part III, and that where there is no likelihood of the trial being completed within a reasonable time, the rigour of such provisions will melt down.45 The Court applied the same reasoning to a prosecution under the PMLA in Manish Sisodia v. Directorate of Enforcement, holding that prolonged incarceration without the prospect of a speedy trial violated Article 21 and reiterating that bail is the rule even under special statutes.46
These decisions matter, but they operate only after long detention and only for those able to reach the Supreme Court. The reversed presumption remains the starting point in the trial court. An accused person under these statutes therefore experiences the rule of bail as something to be earned through years of custody rather than a right available at the threshold.
Judicial responses: from Hussainara to Satender Kumar Antil
The Supreme Court has repeatedly tried to manage the problem through structural directions. Hussainara Khatoon led to the release of thousands of under-trials in Bihar and placed the question of prison populations firmly within constitutional litigation.47 In Bhim Singh v. Union of India, the Court directed jurisdictional magistrates and Sessions Judges to hold sittings in jails to identify and release under-trials eligible under section 436A.48 In the long-running prison conditions case, it required Under-Trial Review Committees to be set up in every district to review the cases of under-trials periodically.49
The most comprehensive intervention is Satender Kumar Antil v. CBI. The Court classified offences into categories for the purpose of bail, directed that the conditions in Arnesh Kumar be followed strictly, held that courts need not insist on arrest at the stage of taking a charge-sheet on record where the accused has cooperated with the investigation, and called on the Union Government to consider a separate law on bail, in the manner of the Bail Act in the United Kingdom.50 It also directed that bail applications be decided within two weeks and anticipatory bail applications within six weeks, save in exceptional cases. In Siddharth v. State of U.P., the Court had earlier made clear that arrest is not a formality to be completed merely because a charge-sheet is filed.51
These are serious efforts, and they have produced results in particular cases. Yet they share a weakness. Each depends on a large number of officials, from police officers to jail superintendents to trial judges, changing their behaviour in response to directions they may never read and are seldom held accountable for ignoring. The Supreme Court can declare the rule, but it cannot sit in every remand court.
The Under-Trial Review Committees illustrate the point. Where they meet regularly and are supported by the District Legal Services Authority, they have helped identify prisoners eligible for release and moved applications on their behalf. Where they meet irregularly, they become a reporting exercise.52 Periodic campaigns by legal services authorities have produced releases in large numbers, but the need for repeated campaigns shows that the ordinary system does not release eligible prisoners on its own.53 A right that depends on a campaign to be enforced has not yet become part of the routine of the courts.
Towards a workable liberty
If the failure is one of design, the response must also be one of design. Six changes would make the rule of bail more than a slogan.
First, Parliament should enact a dedicated Bail Act, as the Supreme Court suggested in Satender Kumar Antil.54 Such a law should state a statutory presumption in favour of release, list exhaustively the grounds on which bail may be refused, require written reasons for refusal, and set time limits for deciding applications. A single statute would also make it easier for trial courts to apply the law consistently.
Second, the use of monetary sureties should be curtailed. For most offences, release on a personal bond with non-monetary conditions should be the default, and any surety amount should be fixed only after an inquiry into the accused’s means. The recommendations of the Law Commission’s 268th Report provide a ready basis for such a reform.55 Where an accused person has been granted bail but remains in custody for a week because of inability to furnish surety, the court should be required to reconsider the conditions on its own motion.
Third, legal representation at the first production must be real. A legal aid lawyer should be present at every first production and remand hearing, should have time to speak to the accused, and should be able to move for bail at once. Strengthening the public defender model under the National Legal Services Authority is the most direct way of achieving this.56
Fourth, the statutory ceilings in section 479 of the BNSS should be made self-executing. The jail superintendent’s duty to apply should be backed by a digital alert in the prison management system when a prisoner crosses one-third or one-half of the maximum sentence, and Under-Trial Review Committees should report compliance publicly.57 The modest use of section 436A shows that a right without a responsible official is easily forgotten.58
Fifth, the special statutes need a time-bound safety valve. Where the trial under the UAPA, the PMLA or the NDPS Act has not concluded within a fixed period, the reversed presumption should cease to apply and the ordinary law of bail should govern. This would give legislative form to what K.A. Najeeb and Manish Sisodia have already recognised as a constitutional requirement.59
Sixth, the system should measure what it does. Courts should publish data on the number of bail applications received, the time taken to decide them, the proportion granted and the number of persons who remain in custody after bail is granted. The comparison with the United Kingdom is useful. Section 4 of the Bail Act 1976 sets out a general right to bail, and Schedule 1 to that Act lists the limited grounds on which it may be withheld, so that every refusal can be tested against a known standard.60 India need not copy that model, but it should adopt its central idea: a decision about liberty must be made against stated criteria and must be visible to those who review it.
Conclusion
The law of bail in India suffers from a strange paradox. Its principles are among the most liberal in the common law world, and its courts have spoken eloquently of the value of personal liberty for almost fifty years. Its prisons, meanwhile, are filled with persons the law presumes to be innocent, many of them young, poor and without effective representation. The paradox is resolved once one sees that the principle and the practice are produced by different actors. The principle is written in appellate judgments. The practice is made in police stations, remand courts and jail offices.
“Bail, not jail” will remain an aspiration until the institutions that make the practice are reorganised around it. The Bharatiya Nagarik Suraksha Sanhita has added useful tools, such as section 479, but it has also added new reasons for caution. A dedicated bail law, the decline of money bail, real legal aid at the first hearing and enforceable time limits would not change the principle. They would simply allow it to be applied, which is, after all, what a constitutional right is supposed to mean.
*****
Footnotes
1. State of Rajasthan v. Balchand, (1977) 4 SCC 308 (India).
2. National Crime Records Bureau, Prison Statistics India 2024 tbl. 2.1 (Ministry of Home Affairs, Government of India, May 2026).
3. National Crime Records Bureau, Prison Statistics India 2022 ch. 2, chart 2.12 (Ministry of Home Affairs, Government of India); National Crime Records Bureau, Prison Statistics India (2005–2024 eds.).
4. Helen Fair & Roy Walmsley, World Pre-trial/Remand Imprisonment List 2 (5th ed. 2025), https://www.prisonstudies.org/sites/default/files/publications/world_pre-trial-remand_imprisonment_list_5th_edition.pdf.
5. India Const. art. 21.
6. Maneka Gandhi v. Union of India, (1978) 1 SCC 248 (India).
7. India Const. art. 22(1)–(2).
8. India Const. art. 39A.
9. D.K. Basu v. State of W.B., (1997) 1 SCC 416 (India).
10. Gudikanti Narasimhulu v. Public Prosecutor, High Court of Andhra Pradesh, (1978) 1 SCC 240 (India).
11. Moti Ram v. State of M.P., (1978) 4 SCC 47 (India).
12. Sanjay Chandra v. CBI, (2012) 1 SCC 40 (India).
13. International Covenant on Civil and Political Rights art. 9(3), Dec. 16, 1966, 999 U.N.T.S. 171. India acceded to the Covenant on Apr. 10, 1979.
14. The Code of Criminal Procedure, 1973, § 436, No. 2, Acts of Parliament, 1974 (India); now The Bharatiya Nagarik Suraksha Sanhita, 2023, § 478, No. 46, Acts of Parliament, 2023 (India).
15. The Code of Criminal Procedure, 1973, §§ 437, 439, No. 2, Acts of Parliament, 1974 (India); now The Bharatiya Nagarik Suraksha Sanhita, 2023, §§ 480, 483, No. 46, Acts of Parliament, 2023 (India).
16. The Code of Criminal Procedure, 1973, § 167(2), No. 2, Acts of Parliament, 1974 (India); now The Bharatiya Nagarik Suraksha Sanhita, 2023, § 187(3), No. 46, Acts of Parliament, 2023 (India).
17. Bikramjit Singh v. State of Punjab, (2020) 10 SCC 616 (India); M. Ravindran v. Intelligence Officer, Directorate of Revenue Intelligence, (2021) 2 SCC 485 (India).
18. The Code of Criminal Procedure, 1973, § 436A, No. 2, Acts of Parliament, 1974 (India), inserted by The Code of Criminal Procedure (Amendment) Act, 2005, No. 25, Acts of Parliament, 2005 (India).
19. The Bharatiya Nagarik Suraksha Sanhita, 2023, § 479, No. 46, Acts of Parliament, 2023 (India).
20. In re Inhuman Conditions in 1382 Prisons, Writ Petition (Civil) No. 406 of 2013, ¶¶ 3–4 (Aug. 23, 2024) (order) (India).
21. Arnesh Kumar v. State of Bihar, (2014) 8 SCC 273 (India).
22. The Bharatiya Nagarik Suraksha Sanhita, 2023, § 35, No. 46, Acts of Parliament, 2023 (India).
23. The Bharatiya Nagarik Suraksha Sanhita, 2023, § 187(2)–(3), No. 46, Acts of Parliament, 2023 (India).
24. PRS Legislative Research, The Bharatiya Nagarik Suraksha Sanhita, 2023: Key Issues and Analysis, https://prsindia.org/billtrack/the-bharatiya-nagarik-suraksha-sanhita-2023 (last visited Nov. 13, 2026).
25. Prison Statistics India 2024, supra note 2, tbls. 6.4–6.5.
26. Prison Statistics India 2024, supra note 2, chart 2.8, tbl. 2.15.
27. Prison Statistics India 2024, supra note 2, tbls. 2.11A, 2.11C, 2.11D.
28. Prison Statistics India 2022, supra note 3, tbls. 7.3–7.4.
29. Prison Statistics India 2024, supra note 2, tbls. 1.2, 1.5, chart 2.3.
30. Law Commission of India, Report No. 268, Amendments to Criminal Procedure Code, 1973: Provisions Relating to Bail ¶¶ 10.1, 11.10–11.12 (May 2017).
31. Law Commission of India, Report No. 78, Congestion of Under-trial Prisoners in Jails (1979).
32. Ministry of Home Affairs, Government of India, Support to Poor Prisoners Scheme (2023).
33. National Legal Services Authority, Legal Aid Defense Counsel System Scheme (2022).
34. Manish Sisodia v. Directorate of Enforcement, 2024 SCC OnLine SC 1920, ¶ 53 (India).
35. Hussainara Khatoon v. Home Secretary, State of Bihar, (1980) 1 SCC 81 (India).
36. P. Ramachandra Rao v. State of Karnataka, (2002) 4 SCC 578, ¶ 20 (India); India Justice Report: Ranking States on the Capacity of Police, Judiciary, Prisons and Legal Aid 79 (2025).
37. Report No. 268, supra note 30, ¶ 4.14.
38. In re Policy Strategy for Grant of Bail, Suo Motu Writ Petition (Criminal) No. 4 of 2021 (Jan. 31, 2023) (order) (India).
39. The Unlawful Activities (Prevention) Act, 1967, § 43D(5), No. 37, Acts of Parliament, 1967 (India).
40. National Investigation Agency v. Zahoor Ahmad Shah Watali, (2019) 5 SCC 1 (India).
41. The Prevention of Money-Laundering Act, 2002, § 45, No. 15, Acts of Parliament, 2003 (India).
42. Nikesh Tarachand Shah v. Union of India, (2018) 11 SCC 1 (India).
43. Vijay Madanlal Choudhary v. Union of India, 2022 SCC OnLine SC 929 (India).
44. The Narcotic Drugs and Psychotropic Substances Act, 1985, § 37, No. 61, Acts of Parliament, 1985 (India).
45. Union of India v. K.A. Najeeb, (2021) 3 SCC 713 (India).
46. Manish Sisodia, supra note 34, ¶¶ 49–57.
47. Hussainara Khatoon, supra note 35.
48. Bhim Singh v. Union of India, (2015) 13 SCC 605 (India).
49. In re Inhuman Conditions in 1382 Prisons, (2016) 3 SCC 700 (India).
50. Satender Kumar Antil v. CBI, (2022) 10 SCC 51 (India).
51. Siddharth v. State of U.P., (2022) 1 SCC 676 (India).
52. India Justice Report, supra note 36, at 133.
53. In re Policy Strategy for Grant of Bail, supra note 38.
54. Satender Kumar Antil, supra note 50.
55. Report No. 268, supra note 30, ¶¶ 11.10–11.12.
56. National Legal Services Authority, supra note 33.
57. The Bharatiya Nagarik Suraksha Sanhita, 2023, § 479(3), No. 46, Acts of Parliament, 2023 (India).
58. Prison Statistics India 2022, supra note 3, tbl. 7.4.
59. K.A. Najeeb, supra note 45; Manish Sisodia, supra note 34.
60. Bail Act 1976, c. 63, § 4, sch. 1 (UK).
