Inside India’s Custodial Violence Crisis

As the National Human Rights Commission records nearly 2,740 custodial deaths in a single year, systemic failures rooted in colonial-era policing, political interference, and institutional impunity continue to erode constitutional guarantees. A landmark Supreme Court verdict issued nearly two decades ago remains largely unimplemented.

Every day, on average, more than six people die in some form of State custody in India. This is not a statistic buried in activist reports; it is drawn from official data submitted before Parliament and compiled by the National Human Rights Commission (NHRC). In 2021–22, the NHRC recorded 2,152 deaths in judicial custody and 155 deaths in police custody, a combined figure that translates to roughly 6.3 custodial deaths per day. By 2023, that number had climbed to approximately 2,400, and in 2024, the NHRC reported 2,739 custodial deaths, the highest figure in recent memory. Behind each number is a person, often poor, often from a marginalised community, often an undertrial who had not yet been convicted of any crime, who entered State custody alive and did not leave that way.

India's custodial death data, when examined in aggregate, reveals a deeply troubling pattern. According to the National Crime Records Bureau (NCRB), an average of 92 deaths occurred in police custody alone every year between 2000 and 2022. The highest single-year police custody toll was 128 in 2005; the most recent years have seen figures hovering between 100 and 175. But police custody represents only a fraction of the full picture. Judicial custody prisons and remand facilities account for the overwhelming majority of deaths. Between 2016 and 2022, a total of 11,419 custodial death cases were recorded by the NHRC across both forms of custody. Of these, compensation was recommended in a mere 1,184 cases, roughly 10 per cent and disciplinary action was initiated in just 21 cases, accounting for a staggering 0.23 per cent of all reported deaths. Geographically, the burden is unevenly distributed. Between April 2018 and March 2023, Gujarat reported the highest number of police custody deaths at 81, followed by Maharashtra with 80, Madhya Pradesh with 50, and Bihar with 47. In judicial custody, Uttar Pradesh consistently registers the highest absolute numbers, 448 deaths in 2021–22 alone and 2,528 over the six years from 2016 to 2022. India's prison system adds another layer to this crisis. The NCRB's Prison Statistics India 2022 report documented 1,995 deaths in judicial custody in that year alone, including 159 officially classified as "unnatural." The national prison occupancy rate has breached 131.4 per cent of designed capacity, meaning that tens of thousands of inmates are confined in conditions that courts and human rights bodies have repeatedly described as incompatible with dignity. Of all those in prison, over 70 per cent are undertrials persons who have not been convicted but are awaiting trial, sometimes for years.

Custodial violence encompasses any act of physical torture, mental harassment, illegal interrogation, humiliation, or threats visited upon a person while in police or judicial custody. When such violence results in death, it becomes a custodial death, an event that triggers specific constitutional and statutory obligations, obligations that are routinely ignored.

The causes of custodial death, as classified by official data, reveal their own uncomfortable story. According to NHRC and NCRB records, approximately 69 per cent of police custody deaths between 2010 and 2020 were attributed to illness (40 per cent) or suicide (29 per cent). Physical assault by police was officially recorded in only 6 per cent of cases. Rights groups and legal scholars have long argued that the "illness" and "suicide" categories obscure a far grimmer reality, injuries sustained during interrogation that go untreated, or psychological collapse induced by abuse that the system then records as natural causes.

Suicides, the data shows, constituted 74.8 per cent of unnatural deaths in prisons in 2022 and 64 per cent in 2023. Whether those deaths represent genuine self-harm or the final act of persons driven to despair by torture and isolation is a question that magisterial inquiries rarely answer with conviction. India's constitutional architecture offers robust, at least on paper, protections against custodial abuse. Article 21 of the Constitution guarantees the right to life and personal liberty, which the Supreme Court has interpreted expansively to include the right not to be subjected to torture, degrading treatment, or arbitrary deprivation of liberty. Article 20(3) protects against self-incrimination, rendering confessions obtained through coercion constitutionally invalid. Sections 24 and 25 of the Indian Evidence Act, now mirrored in the Bharatiya Sakshya Adhiniyam, 2023, make confessions to the police and confessions obtained through inducement, threat, or promise inadmissible before a court. The criminal procedure law is equally explicit. Section 176(1A) of the Code of Criminal Procedure, carried forward as Section 196(3) of the Bharatiya Nagarik Suraksha Sanhita, 2023, mandates that a judicial magistrate, not a fellow police officer, must conduct an inquiry into every death in custody, every disappearance from custody, and every instance of rape in custody. The Jharkhand High Court, in proceedings currently ongoing before it, has found that even this basic obligation is routinely flouted. A Division Bench led by Chief Justice M.S. Sonak, hearing a Public Interest Litigation after state data revealed over 450 custodial deaths in the state between 2018 and 2025, has noted that the Home Department's own affidavit fails to confirm that mandatory judicial magistrate inquiries were conducted in each case.

Sections 330 and 331 of the Indian Penal Code, now Sections 116 and 117 of the Bharatiya Nyaya Sanhita, 2023, specifically criminalise causing hurt or grievous hurt to extort confessions or information, with enhanced punishment when committed by a public servant. Prosecutions under these provisions remain vanishingly rare. India signed the United Nations Convention against Torture in 1997 but has not ratified it, one of the few democracies in the world to remain in this legal limbo for nearly three decades. In September 2006, the Supreme Court of India delivered its judgment in Prakash Singh and Others v. Union of India and Others, a case that had originated a decade earlier as a writ petition filed by Prakash Singh, a retired IPS officer who had served as Director General of Police of both Uttar Pradesh and Assam. The Court, invoking Articles 32 and 142 of the Constitution, acknowledged what decades of high-powered committees the National Police Commission of 1977–81, the Ribeiro Committee, the Padmanabhaiah Committee, the Malimath Committee, and finally the Sorabjee Committee of 2005–06 had already documented: that the Indian Police Act of 1861, a law designed to serve colonial administration rather than a democratic republic, had never been replaced, and that pervasive political interference in policing had produced chronic dysfunction, illegal detentions, custodial torture, fabricated cases, and selective enforcement. The Court issued seven binding directives to all states and Union Territories, to be implemented by December 31, 2006. These included the constitution of a State Security Commission in each state to insulate policing from political interference; a fixed minimum tenure of two years for Directors General of Police, selected from a panel of three senior-most officers prepared by the Union Public Service Commission; similar tenure protections for officers at the Inspector-General level and above; the constitution of a Police Establishment Board to handle transfers and postings of officers below the DGP rank; the creation of a Police Complaints Authority at state and district levels to receive and act on public complaints of abuse and misconduct; the separation of investigation functions from law-and-order duties; and the constitution of a National Security Commission at the Central level for Central Police Organisations. The judgment was, in the assessment of constitutional lawyers, path-breaking not only for its content but for its method. The Supreme Court had, in the absence of legislative will, issued binding executive directions to cure a fundamental failure of governance. Almost two decades later, the failure to implement those directions is itself a case study in institutional resistance. As of 2020, no state was fully compliant with the Supreme Court's directives. Eighteen states had passed or amended Police Acts in the intervening years, but none matched the legislative models the Court had recommended. Multiple contempt petitions have been filed. The Court's amicus curiae in the case, Senior Advocate Raju Ramachandran, submitted an application noting plainly that "the Court's directions in the judgment are unimplemented or not fully implemented." Large states Maharashtra, Tamil Nadu, and Uttar Pradesh have been among the worst performers. State laws were, in the assessment of reformers, "deliberately formulated to give legal garb to the status quo that existed before."

The persistence of custodial torture and death, despite constitutional provisions, Supreme Court directives, and mandatory inquiry requirements, flows from a convergence of structural incentives and institutional failures. The pressure on police to show "quick results" arrests, confessions, and closed cases without the forensic infrastructure or investigative training to achieve them through lawful means creates a powerful incentive for coercive interrogation. Third-degree methods, as they are euphemistically called, have been normalised across generations of officers who were themselves trained in their use and who operate in an environment where accountability is the exception rather than the rule. The colonial inheritance is not merely a rhetorical point. The Indian Police Act of 1861 was designed to give a foreign administration tight control over a subject population. Its basic architecture, the concentration of disciplinary and investigative authority within the police hierarchy, the absence of civilian oversight, and the primacy of the executive over operational decisions survived independence largely intact and continue to shape institutional culture. Marginalised communities bear a disproportionate share of the harm. Dalits, Adivasis, Muslims, migrant labourers, and those without economic or social capital to access legal representation are far more likely to experience custodial violence than those with resources to navigate the system. The 2025 Global Torture Index Factsheet on India notes that incidents involving severe beatings and forced confessions "particularly target marginalised communities such as Dalits, Adivasis, Muslims, LGBTQIA+ individuals and migrant labourers." The structure of impunity reinforces itself. Officers who commit custodial violence are investigated when they are investigated at all by colleagues within the same department, creating a conflict of interest that courts have repeatedly noticed and that the new mandatory judicial magistrate inquiry requirement was designed, at least in part, to address. The weakness of witness and victim protection means that survivors and families who attempt to complain face harassment and, in some cases, prosecution in retaliatory cases. Complaints authorities, where they exist, are under-resourced, slow, and lack enforcement powers.

The Supreme Court's 2006 directives remain the most authoritative and comprehensive blueprint for police reform that India has produced. Their genuine implementation, not the cosmetic version that some states have substituted, remains the most urgent first step.

Beyond the Prakash Singh framework, there is a growing consensus among former police officers, legal scholars, and human rights practitioners on several additional measures. Modern forensic training, including cyber-investigation capabilities, would reduce dependence on confession-based policing. Mandatory installation and real-time monitoring of CCTV cameras in all police stations and remand facilities would create a contemporaneous record that makes it far more difficult to cover up violence. Body-worn cameras for officers engaged in arrests and interrogations are already deployed in several jurisdictions worldwide and have demonstrably reduced complaint rates. Independent medical examination of all persons entering custody, not by police-affiliated doctors but by independent medical officers, would help establish baselines against which any subsequent injuries must be explained. Protection for officers who refuse unlawful orders and for those who report abuse by colleagues, a category that the current system punishes rather than protects, is essential if cultural change is to follow structural change. The ratification of the UN Convention against Torture, pending since 1997, would bring India's legal obligations in line with its stated commitments and create an additional accountability mechanism through the international reporting cycle. The 2020 deaths of P. Jayaraj and his son J. Bennix in police custody in Tamil Nadu's Sathankulam town were, in their horror, also an illustration of what meaningful judicial intervention can look like. The two men died within days of each other after being taken into custody by police on a trivial allegation. Medical evidence documented extensive injuries. The Madras High Court intervened directly, the investigation was transferred to the Central Bureau of Investigation, and several police personnel were arrested, an outcome that remains rare but demonstrates that accountability is not merely theoretical.The case galvanised public attention and renewed demands, across party lines, for a standalone anti-torture law and for the complete implementation of the Prakash Singh directives. Neither has materialised in the years since.Courts have said, with clarity and consistency, that crime control does not justify torture. In a long line of judgments from Nilabati Behera v. State of Orissa (1993), which established the principle of State liability for custodial deaths, to D.K. Basu v. State of West Bengal (1997), which codified detailed procedural safeguards for arrest and detention the Supreme Court has made it plain that the police operate under the law and not above it, and that ends do not justify illegal means.

Article 21's guarantee of life and personal liberty is not suspended at the threshold of a police station or a prison gate. The person in custody, whether an accused, an undertrial, or a convicted prisoner, retains constitutional personhood. To subject that person to torture is not merely a criminal act under the Bharatiya Nyaya Sanhita; it is a constitutional violation that engages the liability of the State itself. Police reform, in this frame, is not a question of administrative preference or budgetary priority. It is a question of whether the Indian State means what its Constitution says.Custodial deaths recorded by NHRC (2024): approximately 2,739, the highest figure in recent years, translating to roughly 7.5 per day.Custodial deaths recorded by NHRC (2021–22): 2,307 total, 2,152 in judicial custody, 155 in police custody.Police custody deaths (NCRB annual average, 2000–2022): 92 per year.Total custodial death cases recorded by NHRC (2016–2022): 11,419.Cases resulting in disciplinary action (2016–2022): 21, or 0.23 per cent of the total.Prison occupancy rate, national average (NCRB, 2022): 131.4 per cent.Undertrials as proportion of prison population: over 70 per cent.States fully compliant with the Supreme Court's Prakash Singh directives (as of 2020): zero.Years since India signed the UN Convention against Torture without ratifying it: 28.

The LB editorial policy requires that reports on custodial deaths be accompanied by the following note: Persons with concerns about treatment in police or judicial custody may approach the National Human Rights Commission (nhrc.nic.in), the State Human Rights Commission, or a legal aid authority under the National Legal Services Authority (nalsa.gov.in). Legal representation at the time of arrest is a right, not a privilege.

https://video.wixstatic.com/video/1fb4ae_d636585f37d04b49816bda24c50f1463/720p/mp4/file.mp4