There's a principle Indian courts have repeated for decades: bail is the rule, jail is the exception. In practice, though, lower courts rarely apply it that way. And when that happens, pre-trial detention stops functioning as a safeguard and starts working like a tax one levied specifically on people who can't afford to pay their way out of it. For thousands of undertrials, incarceration begins long before guilt is ever established in court, and for many, it never gets tested at all.
The latest National Crime Records Bureau (NCRB) Prison Statistics India report puts numbers to this. On paper, the news looks almost reassuring: national prison occupancy has dropped to 112.7%. But dig one level deeper, and the picture turns grim. Of the 5,11,542 people currently in Indian jails, 72.6% haven't been convicted of anything.
Some states are in far worse shape than the national average suggests. Delhi tops the list at a staggering 194.6% occupancy, with undertrials making up more than 87% of that number. Meghalaya isn't far behind at 163.5%, followed by Jammu & Kashmir (148.3%), Madhya Pradesh (147.1%), and Maharashtra (143.9%).
Who exactly is caught in this backlog?
Mostly young people 86.3% of prisoners fall between 18 and 50 years old. Education levels tell their own story too: nearly 68% of inmates are either illiterate (26.2%) or never made it past Class 10 (41.6%). And the social pattern is just as stark Scheduled Castes (21.8%), Scheduled Tribes (14.1%), and Other Backward Classes (33.2%) together account for close to 69% of everyone behind bars. In other words, India's undertrial crisis falls hardest on communities that already carry the weight of historical disadvantages.
Judicial intervention: Statutory safeguards under Section 41 and the Arnesh Kumar mandate
None of this is happening in a legal vacuum India has fairly strong statutory protections against exactly this kind of overreach. The problem is enforcement.
In Arnesh Kumar v. State of Bihar (2014), a Supreme Court Bench of Justices Chandramauli Kr. Prasad and Pinaki Chandra Ghose addressed mechanical arrests and perfunctory judicial remands.
Targeting offences punishable by up to seven years of imprisonment, the Court held that the existence of police power to arrest does not justify its routine exercise, emphasising that detention must remain an exceptional measure under Article 21.
As the Court pointedly observed, "The existence of the power to arrest is one thing, the justification for the exercise of it is quite another... The attitude to arrest first and then proceed with the rest is despicable."
The legal bedrock relies on Section 41(1)(b) CrPC (now Section 35 BNSS).
The Court ruled that police cannot arrest merely upon receiving a complaint. Officers must explicitly record factual satisfaction that arrest is necessary to prevent further crime, enable proper investigation, stop evidence tampering, prevent witness intimidation, or ensure court attendance.
Where custody is unnecessary, Section 41A CrPC obligates serving a Notice of Appearance to allow the accused to join the investigation without arrest.
The legal promise: Constitutional benchmarks
The roots of this fight go back further, to one of India's earliest and most consequential public interest cases: Hussainara Khatoon & Ors. v. Home Secretary, State of Bihar (1979) exposed this systemic cruelty. Sparked by a series of investigative reports by K F Rustamji in a national daily, advocate Pushpa Kapila Hingorani filed India’s first reported Public Interest Litigation.
The petition laid bare a harrowing reality: over 40,000 undertrials were rotting across Bihar’s prisons in facilities like Patna, Muzaffarpur, and Ranchi Central Jails for periods far exceeding the maximum prison terms. Among them was Hussainara Khatoon, a woman held in "protective custody" for over four years without a trial.
Ruling on the petition, a Supreme Court Bench led by Justice P N Bhagwati recognized speedy trial and free legal aid as fundamental rights under Article 21.
Heavily criticizing an archaic legal system that conditions liberty on financial leverage, Justice Bhagwati observed that the poor experience a "law for the poor" rather than a "law of the poor", turning bail into an instrument of economic oppression.
The court directed that indigent accused be released on Personal Recognizance (PR) bonds without demanding monetary sureties or property titles.
More than four decades later, in Satender Kumar Antil v. CBI (2022), the Supreme Court had to confront the same failure all over again this time among subordinate courts still defaulting to custody as a matter of habit.
The court sorted offences into four categories (A through D) to remove ambiguity from judicial discretion. For Category A offences those carrying sentences of seven years or less the ruling was direct: bail should generally be granted without requiring custody first, and notices of appearance should be the norm rather than the exception.
The court went further, directing that if someone is granted bail but can't arrange the required sureties or financial security within seven days, the trial court must step in on its own and convert those conditions into a personal bond.
Frustrated by how routinely trial courts still reach for custody, the judges called on Parliament to draft a dedicated Bail Act something along the lines of the UK's Bail Act, 1976 to take the guesswork and inconsistency out of the process.
Maximum detention, minimum process: The breakdown of Section 436A
Perhaps the clearest example of a law existing on paper but not in practice is Section 436A of the CrPC, now carried forward and expanded as Section 479 of the Bharatiya Nagarik Suraksha Sanhita (BNSS), 2024.
The rule is straightforward: anyone accused of an offence (other than those punishable by death or life imprisonment) who has spent half the maximum possible sentence for that offence in detention, whether during investigation, inquiry, or trial must be released on a personal bond.
Under the new BNSS provision, first-time offenders get an even stronger protection, qualifying for release after serving just one-third of the maximum period.
To actually make this happen rather than let it sit unused, the Supreme Court in Bhim Singh v. Union of India (2014) issued specific, time-bound orders. A bench led by then-Chief Justice R M Lodha directed Judicial Magistrates, Sessions Judges, and Chief Judicial Magistrates to hold weekly jail-visit court sessions starting October 1, 2014 going into prisons themselves to identify eligible undertrials and issue release orders on the spot, with High Courts checking in on compliance every quarter.
From punitive incarceration to risk-based release: What India can learn from global reform
If India wants to actually decongest its prisons, the fix isn't more punishment it's a shift toward risk-based release, and there are working examples elsewhere to draw from.
In the United States, several jurisdictions have moved toward algorithmic risk assessments instead of cash bail, showing that it's possible to stop detaining people simply because they're poor without putting public safety at risk.
European countries have leaned on community supervision and administrative fines for minor offences, keeping petty cases from clogging up jail systems altogether.
Uganda has taken a different but equally practical approach, setting up "ChainLinked Committees" that force police, magistrates, and prison officials to sit down together and jointly review pending cases, breaking the administrative silos that let files simply sit for years.
For India, the way forward looks fairly clear: stop tying bail to a person's bank balance, give Under-Trial Review Committees real enforcement power instead of an advisory role, and actually operationalize the automatic-release mechanisms already written into Section 479 BNSS.
Until personal recognizance and genuine risk assessment replace monetary sureties, Article 21's promise of liberty will keep existing mostly on paper while thousands of people who haven't been convicted of anything continue paying for a system that isn't working.