Introduction
The police are the citizen’s first point of contact with the coercive authority of the State, and police custody is the least visible site in which that authority is exercised. According to the Supreme Court of India, violence in custody is a direct attack on both the rule of law and the human dignity protected by Article 21.1 Yet nearly thirty years later, torture by law enforcement personnel during detention remains a common practice in India, not a rarity.
On 1 July 2024, India’s three principal criminal laws were replaced by the Bharatiya Nyaya Sanhita, 2023 (“BNS”), the Bharatiya Nagarik Suraksha Sanhita, 2023 (“BNSS”) and the Bharatiya Sakshya Adhiniyam, 2023 (“BSA”).2 The new Codes were presented as a move away from a colonial criminal law designed to punish people, towards one that delivers justice to citizens.3 Measured against that claim, custodial torture is a natural test case for the new laws. It is arguably the clearest example of the continuing colonial relationship between the police and the public: confessions are extracted through violence, investigations are conducted through intimidation, and the officers involved are shielded by the requirement of prior government sanction for their prosecution, now found in Section 218 of the BNSS.4
There is no disagreement that a problem exists. In March 2026, data presented to the Lok Sabha showed that 176 people died in police custody in 2021–22, 163 in 2022–23, 157 in 2023–24, 140 in 2024–25 and 170 in 2025–26 up to 15 March 2026; in 2025–26, Bihar recorded the largest number of such deaths.5 According to the Status of Policing in India Report 2025, which relies on National Crime Records Bureau data, not one police officer was convicted for any of the 1,107 deaths in police custody recorded between 2011 and 2022.6 India signed the UN Convention against Torture in 1997 but has yet to ratify it.7 The first Global Torture Index, released in 2025 by the World Organisation Against Torture with partner organisations, rated India as a country at high risk of systemic torture.8 The Supreme Court of India has declined to direct Parliament to enact an anti-torture law, holding that a court cannot direct the legislature to legislate.9
This paper addresses three questions. First, how do the relevant constitutional and human rights standards apply to the treatment of persons in custody in India? Second, do the BNS, BNSS and BSA enhance, preserve or diminish the protections against torture by law enforcement personnel? Third, if a large body of judicial decisions already guides the prevention of custodial torture, why has impunity persisted, and what reforms can end it?
The paper follows a doctrinal approach and draws on official data from the National Human Rights Commission (NHRC) and the National Crime Records Bureau (NCRB), answers to parliamentary questions, reports of the Law Commission of India and empirical research. Part II establishes the constitutional and conceptual framework. Part III reviews the judicial response. Part IV assesses the new criminal laws provision by provision. Part V identifies the structural factors that sustain impunity. Part VI reviews India’s international obligations. Part VII discusses police reform, with specific focus on Bihar. Part VIII develops recommendations and concludes.
The central argument of this article is that custodial impunity in India has become institutionalised. It is the product of the combined effect of the way the laws are designed, the incentives the system creates and the absence of mechanisms of accountability. Guidelines alone cannot remedy it. The recent recodification of India’s criminal law gave Parliament the opportunity to convert judicial protections into legal requirements, to define torture and to remove the procedural obstacles to prosecuting those who abuse their power. The new Codes gave statutory form to some of the judicial protections, but they did not define torture and did not remove any of the procedural obstacles to prosecution.
Conceptual and constitutional framework
A. Defining custodial torture
Indian statute law does not define torture. Neither the Indian Penal Code nor the BNS defines the term; the word appears only in the illustrations to the offence of voluntarily causing hurt to extort a confession (Section 120 of the BNS, formerly Section 330 of the Indian Penal Code). The generally accepted definition of torture is found in Article 1(1) of the United Nations Convention against Torture (UNCAT), under which torture means “any act by which severe pain or suffering, whether physical or mental, is intentionally inflicted on a person for such purposes as obtaining from him or a third person information or a confession, punishing him for an act he or a third person has committed or is suspected of having committed, or intimidating or coercing him or a third person, or for any reason based on discrimination of any kind, when such pain or suffering is inflicted by or at the instigation of or with the consent or acquiescence of a public official or other person acting in an official capacity.”10 This paper focuses on three aspects of the UNCAT definition: torture includes mental suffering; torture is defined by its purpose; and torture involves the participation, direct or indirect, of a public official.
The International Covenant on Civil and Political Rights (“ICCPR”), to which India became a party in 1979, prohibits torture and cruel, inhuman or degrading treatment absolutely. It also requires that persons deprived of their liberty be treated with respect for human dignity.11 In its 273rd Report, the Law Commission of India recommended that Indian law adopt a definition of torture based on Article 1 of UNCAT.12
In this study, “custodial torture” means the infliction of pain or suffering on a person held in the custody of law enforcement agencies such as the police, whether after formal arrest, during informal detention before the arrest is officially recorded, or while the person is remanded to police custody. The distinction between formal and informal custody is crucial, because much of the torture committed by law enforcement personnel occurs before the arrest is officially recorded.
B. Article 21 and the right to dignity in custody
Article 21 provides that no person shall be deprived of life or personal liberty except according to procedure established by law.13 Since Maneka Gandhi, that procedure must be right, just and fair, and not arbitrary, fanciful or oppressive.14 According to the Supreme Court in Francis Coralie Mullin, the right to life includes the right to live with human dignity; any form of torture or cruel, inhuman or degrading treatment offends that dignity and is therefore prohibited by Article 21.15 In Sunil Batra, the Court held that a prisoner is not denuded of all his fundamental rights on entering prison; lawful incarceration restricts only those rights that are inconsistent with it.16 The nine-judge bench in Puttaswamy later reaffirmed that human dignity is the foundation on which Part III of the Constitution operates.17
C. Articles 20(3) and 22
Article 20(3) protects a person accused of an offence from being compelled to be a witness against himself.18 In Nandini Satpathy, the Court extended Article 20(3) to police interrogation and held that “compulsion” includes not only physical force but also psychological pressure and an atmosphere of coercion.19 In Selvi, the Court held that the involuntary administration of narco-analysis, polygraph and brain-mapping tests violates Article 20(3) and the right to mental privacy under Article 21.20
Article 22(1) and (2) require that a person taken into custody be informed of the grounds of arrest, be allowed to consult and be defended by a legal practitioner of his choice, and be produced before the nearest magistrate within twenty-four hours.21 These three safeguards, notice, counsel and judicial scrutiny, mark the points at which torture can best be prevented.
D. The constitutional problem of proof
The protection that Article 21 of the Indian Constitution offers is robust in doctrine, yet its enforcement is weak. Torture takes place in confined spaces and is witnessed only by the officers who commit it. The Supreme Court has acknowledged on several occasions that police officers seldom testify against their colleagues, and that courts should not insist on direct ocular evidence in custodial cases, since that would deny victims any means of redress.22 This imbalance of evidence runs through every subsequent part of this paper. Unless the burden of explaining what happened in custody shifts to the custodian, Article 21 will remain primarily a declaration of rights rather than a functional protection.
The judicial response: from compensation to surveillance
The absence of a specific law against torture has left the Supreme Court to build most of India’s anti-torture framework through constitutional adjudication. That jurisprudence developed in four broad phases: recognition of the problem, compensation for victims, procedural guidelines for law enforcement officials, and technological surveillance of police stations. Sub-parts A, B, C and E below trace these phases; sub-part D considers the Court’s parallel response to fake encounters and arbitrary arrest, and sub-part F the limits of judicial guidelines. Although each phase added protections on paper, none has secured accountability for the perpetrators of torture in practice.
A. Recognition: the Bihar cases
Modern custodial jurisprudence began in Bihar. In the Bhagalpur blindings of 1979–80, undertrial prisoners were blinded by police officers using needles and acid. In Khatri v. State of Bihar, the Supreme Court held that the State must provide free legal aid to an accused person when he is first produced before a magistrate, and that the police conduct infringed the right to life.23 The Court also awarded monetary compensation to Rudul Sah, who had remained in prison for over fourteen years after his acquittal, establishing that compensation may be awarded under Article 32 for a violation of Article 21.24
B. Compensation as a public law remedy
Nilabati Behera v. State of Orissa concerned a young man who died in police custody. The Court held that compensation for the violation of a fundamental right is a public law remedy based on strict liability, distinct from damages in tort, and that sovereign immunity is no defence to it.25 The Court also relied on Article 9(5) of the ICCPR, which recognises an enforceable right to compensation for unlawful arrest or detention. Nilabati Behera remains the basis of the compensation orders issued by constitutional courts and the NHRC. The difficulty with this precedent is that compensation is paid by the State, not by the police officer who violated the victim’s rights; compensation is therefore usually treated as the end of the matter rather than as the beginning of a criminal prosecution of the officer.
C. Procedural guidelines: Joginder Kumar and D.K. Basu
In Joginder Kumar v. State of U.P., the Supreme Court held that the existence of a power to arrest does not by itself justify an arrest; the police officer must be able to justify it. The Court also held that an arrested person is entitled to have a friend or relative informed of the arrest.26 In 1996, in D.K. Basu v. State of West Bengal, the Court laid down eleven requirements to be followed in every arrest. Among them, officers must bear clear identification; a memo of arrest must be attested by at least one witness and countersigned by the arrestee; a friend or relative of the arrestee must be informed of the arrest; the arrestee must be medically examined every forty-eight hours; the arrestee may meet a lawyer during, though not throughout, interrogation; and information about the arrest must be displayed on a notice board at the police control room.27 Failure to comply renders the official liable to departmental action and to punishment for contempt of court.
The Court also confronted the problem of proof. In State of M.P. v. Shyamsunder Trivedi, it criticised the lower courts for acquitting police officers on technicalities, and observed that insisting on direct evidence ignores the reluctance of police officers to testify against one another.28 In Sube Singh v. State of Haryana, the Court held that compensation may be awarded where custodial violence is established, but only where the violation is patent and incontrovertible, and not on claims that are false or exaggerated.29
D. Encounters and arrest discipline
The Court applied the same reasoning to extra-judicial killings. In Prakash Kadam, it held that police officers who commit fake encounters should be given the death sentence, treating such killings as falling within the “rarest of rare” category.30 In PUCL v. State of Maharashtra, the Court laid down sixteen requirements for the investigation of police encounters, including registration of a first information report (FIR), an independent investigation by the CID or a police team of another police station, and a magisterial inquiry.31 In Arnesh Kumar v. State of Bihar, the Court curbed automatic arrests for offences punishable with imprisonment of up to seven years, and required magistrates to scrutinise the arresting officer’s reasons before authorising detention.32
The Court has continued to strengthen Article 22(1) in recent years. It has held that where an arrested person is not informed of the grounds of arrest, the arrest is vitiated, and the remand that follows cannot stand.33
E. Surveillance: Paramvir Singh Saini
In Paramvir Singh Saini v. Baljit Singh, the Court directed all States and Union Territories to install CCTV cameras with night vision and audio recording throughout every police station, including lock-ups and interrogation rooms, and directed the Union to do the same in the offices of central agencies such as the CBI, NIA and ED. Recordings are to be preserved for eighteen months or, where the available equipment cannot store them for that long, for the longest period possible and never less than one year. Oversight committees are to be constituted at State and district levels.34
In September 2025, after a newspaper reported some eleven deaths in police custody in Rajasthan over seven to eight months, the Court registered a suo motu case on the absence of functional CCTV cameras in police stations.35 The amicus curiae informed the Court that the Union, sixteen States and three Union Territories had not complied with the judgment in Paramvir Singh Saini. In November 2025, the Court reportedly rebuked the Union and the States for taking the Court “too lightly” and set a final deadline for compliance.36
F. The limits of judicial guidelines
The Court has also recognised the limits of what it can do. In Ashwani Kumar v. Union of India, it declined to direct Parliament to enact a law against torture in line with UNCAT, holding that a court cannot direct the legislature to enact a law.37 The result is a system of guidelines whose breach rarely attracts punishment. Nearly five decades of jurisprudence have produced detailed standards, but those charged with enforcing them, namely magistrates, medical officers, senior police officers and human rights commissions, have not applied them consistently.
The new criminal laws: an appraisal
The BNS, BNSS and BSA codify a number of judicial protections for persons arrested or questioned by the police. However, they leave the substantive law on custodial torture largely as it has stood since 1860, and in at least one respect they reduce the protection available to persons in custody. Table 1 summarises the changes relevant to custodial torture; a discussion of each follows.
| Safeguard | Former law | New law | Assessment |
|---|---|---|---|
| Definition of torture | None (IPC) | None (BNS) | No change |
| Hurt to extort confession | IPC §§ 330–331 | BNS § 120 | Renumbered; no aggravation for custody |
| Confinement to extort confession | IPC § 348 | BNS § 127(8) | Renumbered |
| Arrest information and designated officer | CrPC § 41C | BNSS § 37 | Strengthened: display at every station |
| Handcuffing | No statutory rule | BNSS § 43(3) | New power for listed offences; cuts against judicial limits |
| Police custody period | CrPC § 167: 15 days, first 15 days only | BNSS § 187: 15 days in parts within 40 or 60 days | Diluted |
| Magistrate’s inquiry into custodial death | CrPC § 176(1A) | BNSS § 196(2) | Carried forward; “Judicial Magistrate or Metropolitan Magistrate” becomes “the Magistrate” |
| Sanction to prosecute public servants | CrPC § 197 | BNSS § 218 | Retained, with 120-day deemed sanction |
| Confession to police | Evidence Act §§ 24–26 | BSA §§ 22–23 | Retained |
| Presumption of custodial responsibility | Not enacted | Not enacted | No change |
Table 1: Safeguards against custodial torture under the former and the new criminal laws
A. The Bharatiya Nyaya Sanhita: torture without a name
The BNS does not define or criminalise torture as such. Custodial torture must still be prosecuted through general offences: voluntarily causing hurt or grievous hurt to extort a confession,38 wrongful confinement to extort a confession,39 culpable homicide or murder, and, where relevant, rape by a police officer in custody.40 The offences committed by public servants who disobey the law or fail to record information are also retained.41
Three gaps follow. First, there is no offence that captures mental torture, which UNCAT expressly includes. Second, the punishment for causing hurt to extort a confession is not graded by custody: the law treats a beating by a police officer in a lock-up in the same way as hurt caused by a private person to extract information. Third, the BNS does not impose command or superior responsibility on officers who order, encourage or tolerate torture. A Code that added new offences such as organised crime and terrorism could have added a single offence of torture by a public servant. It did not.
B. The BNSS: safeguards codified and diluted
Arrest. Section 35 retains the conditions for arrest without warrant and the notice of appearance that the Supreme Court in Arnesh Kumar had insisted upon. It adds that a person who is infirm or above sixty years of age may not be arrested for an offence punishable with imprisonment of less than three years without the prior permission of an officer not below the rank of Deputy Superintendent of Police.42 Sections 36 and 38 codify the arrest memo and the right to meet an advocate during interrogation.43 Section 37 requires the State Government to designate a police officer in every district and police station to maintain the names and addresses of arrested persons, which must be prominently displayed.44 Sections 47 and 48 require that the arrested person be told the grounds of arrest and that a relative or friend, and the designated officer, be informed.45 These provisions put D.K. Basu on a statutory footing, which is a real gain: a breach is now a breach of statute, not only of a judicial guideline.
Handcuffing. Section 43(3) allows the police to use handcuffs at arrest and production for listed categories of offenders, including habitual offenders and persons accused of organised crime, terrorism and certain other serious offences.46 Earlier decisions of the Supreme Court had treated handcuffing as presumptively inhuman and allowed it only with judicial permission in exceptional cases. The new provision gives the police a statutory discretion that is broader than the judicial standard.
Medical examination and production. Section 53 requires medical examination of the arrested person, and Section 58 forbids detention beyond twenty-four hours without a magistrate’s order.47 These carry forward existing law. They do not require examination by an independent doctor, nor do they require the magistrate to question the arrestee privately about ill-treatment.
Police custody under Section 187. This is the most significant change. Under Section 167 of the CrPC, as interpreted in CBI v. Anupam J. Kulkarni, police custody could not exceed fifteen days and was available only during the first fifteen days of remand. Section 187 of the BNSS now allows the fifteen days of police custody to be taken in whole or in parts at any time during the first forty or sixty days of the total period of detention, depending on the gravity of the offence.48 The provision reflects the view expressed in CBI v. Vikas Mishra that the rule in Anupam J. Kulkarni required reconsideration.49 In July 2026, in State of Andhra Pradesh v. Suda Suresh Veera Venkata Naga Raju, the Supreme Court read Section 187(2) and (3) as deliberately enlarging the window for police custody, so that custody not exceeding fifteen days in aggregate may be sought in parts during the first forty or sixty days, and held that a magistrate or court may not foreclose that window by fixing an absolute, non-extendable limit on police custody; it nonetheless retained, as a condition of the custody, the audio-visual recording of every interrogation session.50 The decision confirms this reading of the text of Section 187, although the Court saw the wider window as serving the needs of investigation. The practical effect of the provision is that an accused may be returned from judicial to police custody repeatedly over a period of up to two months. Every return to the police station reopens the period of greatest risk. For a person already in judicial custody, the threat of re-remand is itself a source of coercion.
Custodial deaths. Section 196(2) carries forward the requirement, introduced into the CrPC in 2005, that a magistrate hold an inquiry, in addition to any inquiry or investigation by the police, whenever a person dies or disappears, or a woman is alleged to have been raped, in police custody or in other custody authorised by a magistrate or court. The CrPC provision required that inquiry to be held by a Judicial Magistrate or Metropolitan Magistrate; Section 196(2) refers only to “the Magistrate”.51 The provision is sound in principle. Its failure lies in practice, as Part V shows.
Audio-video recording. Section 105 requires audio-video recording of searches and seizures.52 The legislature did not extend this duty to interrogation. That omission is striking, because the Supreme Court had already directed audio-video coverage of interrogation rooms in Paramvir Singh Saini. Parliament could have made that direction a statutory obligation with evidentiary consequences for non-compliance.
Sanction. Section 218 retains the requirement of prior government sanction before a court can take cognisance of an offence alleged to have been committed by a public servant while acting or purporting to act in the discharge of official duty. The new provision requires the government to decide within one hundred and twenty days, failing which sanction is deemed to have been accorded.53 The time limit is an improvement, but the sanction barrier remains, and it still applies to police officers.
Preliminary inquiry. Section 173(3) allows the officer in charge of a police station, with the prior permission of an officer not below the rank of Deputy Superintendent of Police, to conduct a preliminary enquiry within fourteen days for cognizable offences punishable with imprisonment of three years or more but less than seven years, before proceeding with the case.54 Wrongful confinement to extort a confession, punishable with imprisonment of up to three years, falls within this band. In a complaint against police officers, the preliminary enquiry gives the accused institution a lawful means to delay registration.
C. The Bharatiya Sakshya Adhiniyam: an unchanged law of proof
The BSA retains the rule that a confession caused by inducement, threat, coercion or promise is irrelevant, and that a confession to a police officer cannot be proved against the accused, subject to the discovery exception.55 These rules remain important deterrents.
What the BSA did not do is equally important. The Law Commission recommended as early as 1985, and again in 1994 and 2003, that a presumption be added: where a person in police custody suffers injury, the court should presume that the injury was caused by the officer who had custody, unless the officer proves otherwise.56 The proposal responded directly to the problem of proof described in Part II. Despite repeated recommendations, the BSA contains no such presumption. The burden of proving what happened behind the closed door still rests on the victim.
Systemic impunity: anatomy of a failure
Impunity for custodial torture in India is not the result of a single gap. It is produced at every stage of the accountability chain: recording, registration, investigation, sanction, prosecution and remedy. A failure at any one stage is usually enough to end the case.
A. The data gap
The first failure lies in counting. For 2020, the NCRB recorded 76 deaths in police custody, the NHRC 90, and the National Campaign Against Torture, a civil society initiative, 111.57 None of the official sources publishes reliable data on non-fatal torture. NCRB data for 2022 show that 55 per cent of deaths in police custody were of persons not on remand, that is, persons in police custody within the first twenty-four hours of arrest.58 This is the period in which no magistrate has seen the detainee, and it is the period that the new Codes do least to regulate.
B. From death to prosecution: the attrition funnel
The Status of Policing in India Report 2025 traces what happened after deaths in police custody between 2018 and 2022. Cases were registered against police personnel in only about 10 per cent of the deaths. Charge-sheets were filed in about 12 per cent of those cases. There were no convictions.59 In 2022, judicial inquiries, which are mandatory in every case of death in police custody, were ordered in only 35 per cent of cases.60 The parliamentary answer of March 2026, as reported, recorded only one case of disciplinary action in connection with custodial deaths over the five years it covered.61
This pattern reveals that the main barrier is not the absence of an offence but the absence of an independent investigator. The FIR, if registered, is investigated by officers of the same force, often of the same district. The investigating officer depends on the goodwill of the accused officers’ superiors and colleagues. Witnesses are usually other police officers or detainees still under police control. The structural conflict of interest is obvious.
C. The sanction barrier
Section 218 of the BNSS, like Section 197 of the CrPC before it, requires prior sanction to prosecute a public servant for an act done while acting or purporting to act in the discharge of official duty. The Supreme Court has held that sanction is required where the act has a reasonable connection with official duty, even if the officer acted in excess of that duty.62 The Court has also said that it is not the duty of police officers to kill an accused merely because he is a dreaded criminal, and that the position on sanction may be completely different where an encounter was stage-managed.63 Between these two positions lies a wide area of uncertainty. In practice, the question whether a beating during interrogation was “in the discharge of official duty” is litigated for years, and the prosecution often stalls. The new 120-day deemed-sanction rule reduces executive delay, but it does not remove the executive veto.
D. The weak watchdogs
The Protection of Human Rights Act, 1993 gives the NHRC and State Human Rights Commissions power to inquire into custodial violence, but their findings are recommendations, not binding orders.64 Commissions cannot inquire into a matter more than one year after the violation, and the NHRC’s power over the armed forces is limited to seeking a report.65 Its principal remedy is compensation. In 2021–22, the NHRC announced compensation of about ₹4.53 crore in 137 cases of custodial death.66 Compensation without prosecution turns accountability into a transaction: the State pays, and the officer remains in service.
E. Magistrates and doctors
The two professional safeguards that the law places between the detainee and the police, the remand magistrate and the medical officer, frequently fail. Lawyers, judges and doctors interviewed for the Status of Policing in India Report describe magistrates who rarely interact with arrested persons, and medical examinations conducted by doctors without training in forensic medicine, who are less able to recognise signs of torture.67 The law requires both safeguards; it does not require either to be effective.
F. Institutional culture
Finally, impunity is sustained by belief. In the 2025 survey of 8,276 police personnel across seventeen States and Union Territories, 30 per cent said that third-degree methods are justified towards the accused in serious criminal cases, and 9 per cent said they are justified even in petty offences.68 Pressure for quick results, shortage of forensic capacity and investigative training, and public tolerance of “tough” policing all make torture appear efficient. The Law Commission identified these causes three decades ago.69 The new criminal laws, by not addressing investigation by coercion directly, leave the culture undisturbed.
International human rights obligations and the UNCAT question
A. The absolute prohibition
The prohibition of torture is among the few norms of international law that permit no derogation. Article 7 of the ICCPR binds India as a State party, and the Human Rights Committee has explained that the prohibition applies in all circumstances, including a public emergency, that complaints must be investigated promptly and impartially, and that those who violate it must be held responsible.70 The prohibition is widely recognised as a peremptory norm of customary international law, binding even on States that have not ratified UNCAT.
B. What UNCAT would require
UNCAT goes further than the ICCPR by prescribing specific measures. It requires each State party to take effective legislative and other measures to prevent torture, and provides that no exceptional circumstance and no order of a superior may justify it.71 It requires that all acts of torture, including attempts, complicity and participation, be offences under criminal law with penalties reflecting their gravity.72 It requires prompt and impartial investigation wherever there is reasonable ground to believe torture has occurred, protection of complainants and witnesses, an enforceable right to fair and adequate compensation, and exclusion of statements obtained by torture.73 The Optional Protocol adds a system of independent visits to places of detention through national preventive mechanisms.74
Measured against these standards, Indian law falls short in four respects: there is no offence of torture; investigation is not independent; compensation is discretionary and not a statutory right; and there is no independent inspection of police lock-ups.
C. The long non-ratification
India signed UNCAT on 14 October 1997. The Prevention of Torture Bill, 2010 was passed by the Lok Sabha but was referred to a Select Committee of the Rajya Sabha, which found its definition of torture too narrow and its limitation period too short. The Bill lapsed in 2014.75 In 2017, the Law Commission submitted its 273rd Report with a revised draft Prevention of Torture Bill, recommending ratification, a UNCAT-compliant definition, a presumption that injuries in custody were caused by the custodian, and compensation for victims.76 The Report also noted that non-ratification has harmed India’s interests, because foreign courts have refused the extradition of fugitives on the ground of the risk of torture in India.77 During India’s fourth Universal Periodic Review in 2022, several States again recommended ratification; India supported most of these recommendations, but it has not ratified the Convention.78
At the same review, India’s delegation stated that the country’s legal framework provided adequate protection against torture.79 The data in Part V call that assessment into question. The absence of a single conviction for the 1,107 deaths in police custody recorded between 2011 and 2022 is not consistent with any reasonable understanding of the duty to punish.
D. Domestic effect of international norms
Indian courts have used international instruments to give content to fundamental rights where domestic law is silent, provided there is no inconsistency with domestic law.80 Nilabati Behera drew on Article 9(5) of the ICCPR to recognise a right to compensation.81 That interpretive method allows courts to read the UNCAT definition and the duty of independent investigation into Article 21 even before ratification. But judicial incorporation cannot create a new criminal offence or alter the burden of proof in a criminal trial. Those steps require Parliament. The BNS, BNSS and BSA were the obvious vehicle, and the opportunity was not taken.
Police reform and accountability
A. The Prakash Singh framework
The structural reform of the police has been on the national agenda since the National Police Commission, appointed in 1977 to review a police system still governed by the Police Act, 1861, reported between 1979 and 1981.82 In Prakash Singh v. Union of India, the Supreme Court issued seven directions to insulate the police from political interference and to make it accountable. They required a State Security Commission, a minimum tenure and merit-based selection for the Director General of Police, a minimum tenure for officers on operational duty, separation of investigation from law and order, a Police Establishment Board, a National Security Commission, and Police Complaints Authorities at State and district levels.83
For custodial torture, the most important of these directions are the separation of investigation from law and order and the Police Complaints Authorities. The first addresses a cause of torture: investigators who are also burdened with law and order duties lack time, training and forensic support, and resort to coercion. The second addresses impunity: a complaints body independent of the police hierarchy, empowered to inquire into custodial death, grievous hurt and rape in custody.
B. Implementation deficit
Two decades later, implementation remains partial. The Model Police Act, 2006, drafted by a committee appointed by the Union Government, offered a statutory template.84 Many States enacted new police statutes, but these generally adopted the framework selectively and kept oversight bodies under executive control. The Second Administrative Reforms Commission, reporting in 2007, endorsed the separation of investigation and independent complaints mechanisms.85 Subsequent orders of the Supreme Court in Prakash Singh have been confined largely to the appointment of the Director General of Police.86
The new criminal laws are silent on police organisation. That is partly constitutional: “police” is a State subject under Entry 2 of List II of the Seventh Schedule.87 But criminal procedure is in the Concurrent List, and Parliament could have created statutory duties of independent investigation into custodial offences, as it did for judicial inquiries into custodial deaths in 2005.
C. Special reference to Bihar
Bihar holds an unusual place in this history. Many foundational judgments on custodial rights arose from the State: the Bhagalpur blindings in Khatri, the illegal detention in Rudul Sah, and the discipline of arrest in Arnesh Kumar.88 Bihar enacted the Bihar Police Act, 2007 in response to Prakash Singh.89 Yet parliamentary data for 2025–26, up to 15 March 2026, recorded Bihar as having the largest absolute number of deaths in police custody in India.90
Three State-level measures follow from the analysis. First, a functional Police Complaints Authority at State and district levels, chaired by a retired judge and with members from outside the police, should be given jurisdiction over every death, disappearance, rape and grievous hurt in custody. Second, the CCTV directions in Paramvir Singh Saini should be implemented in every police station, with monthly verification by the district oversight committee and public disclosure of compliance. Third, the separation of investigation from law and order should be extended from urban police stations to all police stations, supported by trained investigating officers and adequate forensic capacity. These measures need no constitutional amendment; they need political will and budgetary commitment.
D. Community policing and legitimacy
Police reform is not only about oversight. Policing that relies on coercion is also policing that lacks public cooperation. Investigations that start from evidence, not confession, require witnesses who trust the police, forensic support and time. Community policing initiatives, human rights training and evidence-based interrogation techniques have support within the force itself: in the 2025 survey, 79 per cent of police personnel considered training in human rights, and in evidence-based interrogation techniques, very important.91 Reform should build on that support rather than treat the police as uniformly resistant.
Recommendations and conclusion
A. Recommendations
1. Enact a standalone Prevention of Torture Act. The Act should adopt the UNCAT definition, including mental suffering; create a specific offence of torture by a public servant, with graded punishment where torture causes death or grievous hurt; impose liability on superiors who order or knowingly tolerate torture; and remove the sanction requirement for this offence. The draft Bill annexed to the Law Commission’s 273rd Report provides a ready basis.92
2. Ratify UNCAT and the Optional Protocol. Ratification would bring international review, strengthen India’s position in extradition proceedings, and provide a framework for independent inspection of places of detention.
3. Insert a presumption of custodial responsibility into the BSA. Where a person suffers injury or dies in police custody, the court should presume that the injury was caused by the officer having custody unless the officer proves otherwise, as the Law Commission has repeatedly recommended.93
4. Amend Section 187 of the BNSS. Since the Supreme Court has read Section 187 as deliberately widening the window for police custody, any narrowing of that window must now come from Parliament.94 Police custody should be confined to the first fifteen days of remand or, at minimum, each re-remand to police custody should require written reasons, a fresh medical examination by an independent doctor, and a private hearing of the accused by the magistrate.
5. Make audio-video recording of interrogation a statutory duty. Extend Section 105 of the BNSS to all interrogations, require preservation in line with Paramvir Singh Saini, and provide that unexplained absence of footage permits an adverse inference against the prosecution and the custodian.
6. Create independent investigation. Every death, disappearance, rape or grievous hurt in police custody should be investigated by an agency outside the district police chain of command, such as a State-level unit under the Police Complaints Authority or the State Human Rights Commission.
7. Strengthen remand and medical safeguards. Magistrates should be required by rule to ask every detainee, out of police hearing, whether they have been ill-treated, and to record the answer. Medical examinations should follow the Istanbul Protocol standards for documenting torture.
8. Make compensation a statutory right and recover it from offenders. Compensation should be a right enforceable against the State, in line with Article 14 of UNCAT, with recovery from officers found responsible.95
9. Implement Prakash Singh, beginning with Bihar. Functional State and district Police Complaints Authorities, separation of investigation from law and order in all police stations, and verified CCTV coverage should be time-bound priorities.
10. Publish complete data. The NCRB should publish annual data on custodial torture complaints, FIRs, charge-sheets and convictions against police personnel, disaggregated by State and by the social profile of victims.
B. Conclusion
India’s constitutional law on custodial torture is among the most developed in the common law world. The Supreme Court has recognised that torture violates Article 21, made compensation a public law remedy, prescribed detailed arrest procedures, and ordered surveillance of every police station. Yet the outcome, measured by convictions, is close to zero. That gap between norm and outcome is the clearest evidence that impunity is systemic.
The BNS, BNSS and BSA were presented as a decolonisation of criminal law. On custodial torture, they are better described as a recodification. They give statutory form to parts of D.K. Basu, carry forward the magistrate’s inquiry into custodial deaths and impose a time limit on sanction decisions. But they leave torture undefined, retain the sanction barrier, decline to shift the burden of proof, omit recording of interrogation, and, through Section 187, extend the period during which an accused may be returned to police custody.
A genuinely post-colonial criminal law would treat the police lock-up as the place where the State’s commitment to dignity is tested most severely. It would define torture, punish it specifically, investigate it independently, and presume responsibility where a person enters custody unharmed and leaves it injured or dead. Until Parliament takes those steps, the promise of Article 21 will continue to stop at the door of the police station.
*****
Footnotes
1. D.K. Basu v. State of W.B., (1997) 1 SCC 416, ¶¶ 9–10 (India).
2. The Bharatiya Nyaya Sanhita, No. 45 of 2023, India Code (2023); The Bharatiya Nagarik Suraksha Sanhita, No. 46 of 2023, India Code (2023); The Bharatiya Sakshya Adhiniyam, No. 47 of 2023, India Code (2023). All three came into force on July 1, 2024. See Ministry of Home Affairs, Notifications, Gazette of India, pt. II sec. 3(ii) (Feb. 23, 2024).
3. Dep’t-Related Parliamentary Standing Comm. on Home Affs., Rajya Sabha, Two Hundred Forty-Sixth Report on the Bharatiya Nyaya Sanhita, 2023 ¶ 1.3 (Nov. 10, 2023) (the Sanhita “seeks to change the nature of the law towards providing justice rather than punishment and would be a step forward to remove traces of colonial mind-set”).
4. Bharatiya Nagarik Suraksha Sanhita, § 218(1).
5. Custodial Deaths at 170 This Year, Accountability Nearly Absent: Govt Data, The Wire (Mar. 25, 2026), https://thewire.in/government/custodial-deaths-at-170-this-year-accountability-nearly-absent-govt-data (reporting the answer of the Minister of State for Home Affairs to a Lok Sabha unstarred question of Mar. 24, 2026).
6. Common Cause & Lokniti–CSDS, Status of Policing in India Report 2025: Police Torture and (Un)Accountability 19 (2025), https://www.commoncause.in/wotadmin/upload/SPIR_2025.pdf.
7. Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, Dec. 10, 1984, 1465 U.N.T.S. 85 [hereinafter CAT]. India signed the Convention on Oct. 14, 1997.
8. Azeefa Fathima, First Ever Global Torture Index Flags India as ‘High Risk’ Country, The News Minute (June 26, 2025), https://www.thenewsminute.com/news/first-ever-global-torture-index-flags-india-as-high-risk-country.
9. Ashwani Kumar v. Union of India, (2020) 13 SCC 585 (India).
10. CAT, supra note 7, art. 1(1).
11. International Covenant on Civil and Political Rights arts. 7, 10(1), Dec. 16, 1966, 999 U.N.T.S. 171 [hereinafter ICCPR].
12. Law Comm’n of India, Report No. 273: Implementation of ‘United Nations Convention against Torture and Other Cruel, Inhuman and Degrading Treatment or Punishment’ through Legislation (Oct. 2017).
13. India Const. art. 21.
14. Maneka Gandhi v. Union of India, (1978) 1 SCC 248 (India).
15. Francis Coralie Mullin v. Adm’r, Union Territory of Delhi, (1981) 1 SCC 608, ¶ 8 (India).
16. Sunil Batra v. Delhi Admin., (1978) 4 SCC 494 (India).
17. K.S. Puttaswamy v. Union of India, (2017) 10 SCC 1 (India).
18. India Const. art. 20, cl. (3).
19. Nandini Satpathy v. P.L. Dani, (1978) 2 SCC 424 (India).
20. Selvi v. State of Karnataka, (2010) 7 SCC 263 (India).
21. India Const. art. 22, cls. (1)–(2).
22. State of M.P. v. Shyamsunder Trivedi, (1995) 4 SCC 262 (India); Munshi Singh Gautam v. State of M.P., (2005) 9 SCC 631 (India).
23. Khatri v. State of Bihar, (1981) 1 SCC 627 (India).
24. Rudul Sah v. State of Bihar, (1983) 4 SCC 141 (India).
25. Nilabati Behera v. State of Orissa, (1993) 2 SCC 746 (India).
26. Joginder Kumar v. State of U.P., (1994) 4 SCC 260 (India).
27. D.K. Basu, supra note 1, ¶ 35.
28. Shyamsunder Trivedi, supra note 22.
29. Sube Singh v. State of Haryana, (2006) 3 SCC 178 (India).
30. Prakash Kadam v. Ramprasad Vishwanath Gupta, (2011) 6 SCC 189 (India).
31. People’s Union for Civil Liberties v. State of Maharashtra, (2014) 10 SCC 635 (India).
32. Arnesh Kumar v. State of Bihar, (2014) 8 SCC 273 (India).
33. Vihaan Kumar v. State of Haryana, 2025 SCC OnLine SC 269 (India); see also Prabir Purkayastha v. State (NCT of Delhi), (2024) 8 SCC 254 (India).
34. Paramvir Singh Saini v. Baljit Singh, (2021) 1 SCC 184 (India).
35. In re Lack of Functional CCTVs in Police Stations, Suo Motu Writ Petition (C) No. 7 of 2025 (SC Sept. 4, 2025) (India); see Debby Jain, Supreme Court Takes Suo Motu Case on Lack of CCTV Cameras in Police Stations, LiveLaw (Sept. 4, 2025), https://www.livelaw.in/top-stories/supreme-court-takes-suo-motu-case-on-lack-of-cctv-cameras-in-police-stations-302905.
36. Debby Jain, ‘Why No CCTVs in Interrogation Rooms of Police Stations?’: Supreme Court Asks Rajasthan Govt; Seeks Response on 11 Custody Deaths, LiveLaw (Oct. 14, 2025), https://www.livelaw.in/top-stories/supreme-court-rajasthan-govt-asked-why-no-cctvs-in-police-station-interrogation-rooms-to-respond-on-11-deaths-in-custody-306920; Ritwik Choudhury, “Taking Court Too Lightly”: Supreme Court Raps Centre, States for Failing to Install CCTVs in Police Stations, Bar & Bench (Nov. 25, 2025), https://www.barandbench.com/news/supreme-court-raps-centre-states-for-failing-to-install-cctvs-in-police-stations.
37. Ashwani Kumar, supra note 9.
38. The Bharatiya Nyaya Sanhita, No. 45 of 2023, India Code (2023), § 120; cf. The Indian Penal Code, No. 45 of 1860, India Code (1860), §§ 330–31.
39. Bharatiya Nyaya Sanhita, § 127(8); cf. Indian Penal Code, § 348.
40. Bharatiya Nyaya Sanhita, §§ 64(2)(a), 103, 105.
41. Id. §§ 198–99.
42. The Bharatiya Nagarik Suraksha Sanhita, No. 46 of 2023, India Code (2023), § 35(3), (7).
43. Id. §§ 36, 38.
44. Id. § 37.
45. Id. §§ 47–48.
46. Id. § 43(3); cf. Prem Shankar Shukla v. Delhi Admin., (1980) 3 SCC 526 (India).
47. Bharatiya Nagarik Suraksha Sanhita, §§ 53, 58.
48. Id. § 187(2)–(3); cf. The Code of Criminal Procedure, No. 2 of 1974, India Code (1974), § 167; Central Bureau of Investigation v. Anupam J. Kulkarni, (1992) 3 SCC 141 (India).
49. Central Bureau of Investigation v. Vikas Mishra, (2023) 6 SCC 49, ¶ 7.1 (India).
50. State of Andhra Pradesh v. Suda Suresh Veera Venkata Naga Raju, 2026 INSC 744, ¶¶ 22, 25, 27 (India).
51. Bharatiya Nagarik Suraksha Sanhita, § 196(2); cf. The Code of Criminal Procedure (Amendment) Act, No. 25 of 2005, India Code (2005), § 18 (inserting § 176(1A) into the Code of Criminal Procedure).
52. Bharatiya Nagarik Suraksha Sanhita, § 105.
53. Id. § 218; cf. Code of Criminal Procedure, § 197.
54. Bharatiya Nagarik Suraksha Sanhita, § 173(3).
55. The Bharatiya Sakshya Adhiniyam, No. 47 of 2023, India Code (2023), §§ 22–23; cf. The Indian Evidence Act, No. 1 of 1872, India Code (1872), §§ 24–27.
56. Law Comm’n of India, Report No. 113: Injuries in Police Custody (1985); Law Comm’n of India, Report No. 152: Custodial Crimes (1994); Law Comm’n of India, Report No. 185: Review of the Indian Evidence Act, 1872 (2003).
57. Common Cause & Lokniti–CSDS, Status of Policing in India Report 2025: Executive Summary (2025), https://commoncause.in/wotadmin/upload/Executive_Summary_SPIR_2025.pdf.
58. Id.
59. Id.
60. Id.
61. Custodial Deaths at 170, supra note 5.
62. D. Devaraja v. Owais Sabeer Hussain, (2020) 7 SCC 695 (India).
63. Om Prakash v. State of Jharkhand, (2012) 12 SCC 72, ¶¶ 37–38 (India).
64. The Protection of Human Rights Act, No. 10 of 1994, India Code (1994), § 18.
65. Id. §§ 19, 36(2).
66. Over 2,150 Cases of Deaths in Judicial Custody in 2021–22: Govt Tells Lok Sabha, The Week (Mar. 22, 2022), https://www.theweek.in/news/india/2022/03/22/over-2150-cases-of-deaths-in-judicial-custody-in-2021-22-govt-tells-lok-sabha.html.
67. Common Cause & Lokniti–CSDS, supra note 6, ch. 7.
68. Id. ch. 5; see also NGO Report Claims Gaps by Police Personnel in Adhering to Rule of Law, ThePrint (Mar. 26, 2025), https://theprint.in/india/ngo-report-claims-gaps-by-police-personnel-in-adhering-to-rule-of-law/2566397.
69. Law Comm’n of India, Report No. 152, supra note 56.
70. Hum. Rts. Comm., General Comment No. 20: Article 7 (Prohibition of Torture, or Other Cruel, Inhuman or Degrading Treatment or Punishment), ¶¶ 3, 13–14 (1992).
71. CAT, supra note 7, art. 2.
72. Id. art. 4.
73. Id. arts. 12–15.
74. Optional Protocol to the Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment arts. 3–4, Dec. 18, 2002, 2375 U.N.T.S. 237.
75. The Prevention of Torture Bill, 2010 (as passed by Lok Sabha, May 6, 2010); Rajya Sabha Secretariat, Report of the Select Committee on the Prevention of Torture Bill, 2010 (Dec. 2010).
76. Law Comm’n of India, Report No. 273, supra note 12.
77. Id.
78. Human Rights Council, Rep. of the Working Grp. on the Universal Periodic Review: India, ¶¶ 151.1–151.14, U.N. Doc. A/HRC/52/11 (Dec. 14, 2022); Human Rights Council, Rep. of the Working Grp. on the Universal Periodic Review: India, Addendum, at 2, U.N. Doc. A/HRC/52/11/Add.1 (Feb. 27, 2023) (recording India’s position on each recommendation).
79. Rep. of the Working Grp. on the Universal Periodic Review: India, U.N. Doc. A/HRC/52/11, supra note 78, ¶ 147.
80. Vishaka v. State of Rajasthan, (1997) 6 SCC 241 (India).
81. Nilabati Behera, supra note 25; ICCPR, supra note 11, art. 9(5).
82. Nat’l Police Comm’n, First Report (1979); Prakash Singh v. Union of India, (2006) 8 SCC 1 (India).
83. Prakash Singh, supra note 82.
84. Ministry of Home Affairs, Police Act Drafting Committee, Model Police Act, 2006 (2006).
85. Second Admin. Reforms Comm’n, Fifth Report: Public Order (2007).
86. See, e.g., Prakash Singh v. Union of India, (2019) 4 SCC 1 (India).
87. India Const. sch. VII, list II, entry 2; id. list III, entry 2.
88. Khatri, supra note 23; Rudul Sah, supra note 24; Arnesh Kumar, supra note 32.
89. The Bihar Police Act, No. 7 of 2007, Bihar Code (2007).
90. Custodial Deaths at 170, supra note 5.
91. Common Cause & Lokniti–CSDS, supra note 6, ch. 6.
92. Law Comm’n of India, Report No. 273, supra note 12, annex.
93. See Law Comm’n of India, Report No. 113, supra note 56; Law Comm’n of India, Report No. 273, supra note 12.
94. Suda Suresh, supra note 50, ¶ 22.
95. CAT, supra note 7, art. 14.
