The Script Never Changes. Only the Victims Do.

Amit Patil

Aug 31

17 min read

Custodial Deaths in India: Three Decades of Documented Failure

By Amit A. Patil | Independent Legal Practitioner | Bombay High Court

Estimated reading time: 12 minutes

On April 6, 2026, a sessions court in Madurai sentenced nine Tamil Nadu police officers to death for the custodial murder of a father and his son. The nation called it justice. Lawyers called it something more complicated: a landmark verdict in a country that has been producing landmark verdicts on custodial deaths — without stopping them — for over thirty years.

This essay traces three cases across three decades and three states. It is, ultimately, about one question: why does the most powerful legal system in the world's largest democracy keep documenting this crime without preventing it?

PART I

The Mothers Who Waited

Parvathi, Cuddalore, 1993

There is a phrase that recurs, with eerie consistency, across India's long and troubled record of custodial deaths. A man is taken to a police station. Hours pass, then days. Then comes the announcement, delivered with bureaucratic finality: the detainee has "escaped." No body. No explanation. Just a disappearing act that the authorities expect the family — and the courts — to accept.

In 1993, a tribal labourer named Rajakannu "escaped" from the Kammapuram police station in Cuddalore district, Tamil Nadu. Rajakannu was a member of the Andai Kurumbar tribe, a bamboo-basket maker and agricultural labourer who had been brought in on an allegation that he had stolen jewels from a household that employed him. What happened inside the station was, the police said, nothing unusual. He was there. Then he was not. He had escaped.

His wife Parvathi did not believe it. She had no legal education, no money, and no institutional connection. What she had was the knowledge of a wife: her husband did not run. She walked into the chamber of a young advocate named K. Chandru in Tamil Nadu and asked him to find out what had happened. Chandru took the case.

What he found, painstakingly, over years of litigation, was this: Rajakannu had been beaten to death inside the Kammapuram police station. His body had been secretly transported to a neighbouring district under cover of night and disposed of. Two of his nephews, who had also been held in the same lock-up and who were eyewitnesses to what had happened, had been quietly transferred to a jail in Kerala on fabricated charges — removed from Tamil Nadu so that they could not testify.

Chandru found them anyway. He filed a habeas corpus petition — HCP No. 711 of 1993 — before the Madras High Court. Habeas corpus is among the oldest writs in a common-law system: a demand, in plain Latin, that the State produce the body of a person it is holding and justify that person's detention. In Parvathi's case, there was no body to produce. The question the petition forced the State to answer was why.

There was a procedural obstacle. Witness examination had not traditionally been permitted in habeas corpus proceedings, which were designed as summary remedies — quick judicial checks on arbitrary detention, not full-scale trials. Chandru cited the Kerala High Court's ruling in the Rajan case, where examination of witnesses had been allowed. The Madras High Court accepted the argument. Once the two nephews testified, the State's cover story unravelled.

The 1994 judgment of the Madras High Court ordered a Special Investigation, directed the government to pay compensation to Parvathi, and — in a remarkable order — directed that she be given employment in government service, recognising that the State's crime had not merely killed her husband but destroyed her livelihood. The criminal trial that followed took thirteen more years. The special prosecutor appointed to the case had his house attacked, his library burned, and his car destroyed. He continued. In 2006, the accused police officers were convicted and sentenced to fourteen years' imprisonment.

Parvathi's fight — a poor, illiterate tribal woman against the entire apparatus of the Tamil Nadu State — lasted over a decade. The case inspired the 2021 film Jai Bhim. Justice K. Chandru, who fought it as a young advocate, later became a judge of the Madras High Court, where he disposed of 96,000 cases in six and a half years. His understanding of custodial injustice — the fabricated escapes, the destroyed evidence, the intimidated witnesses, the complicit magistrates — was not learned from textbooks. It was earned in a courtroom in Chennai in 1993.

Asiya Begum, Mumbai, 2002

Nine years later and nine hundred kilometres to the west, a 27-year-old software engineer named Khwaja Yunus came home to Parbhani, Maharashtra, for Eid. He had been working in Dubai. On December 25, 2002, he was arrested in connection with the Ghatkopar bus station bomb blast. Within weeks, the Mumbai Police announced that Yunus had escaped during transfer to Aurangabad. His body has never been found.

His mother, Asiya Begum, went to the courts. A CID inquiry — ordered after a petition by Yunus's father — established that Yunus had died as a result of custodial torture at the Ghatkopar police station. The inquiry named fourteen police personnel. The Maharashtra government granted sanction to prosecute four: Sub-Inspector Sachin Vaze and three constables.

What happened next is a masterclass in institutional impunity. The case was formally admitted for trial only in 2009 — six years after Yunus's death. The special prosecutor who sought to add more accused to the chargesheet was subsequently removed by the state government. The four accused, suspended in 2004, were reinstated into the Mumbai Police in June 2020 — with a murder trial still pending — on the recommendation of a review committee headed by then-Police Commissioner Param Bir Singh. By 2025, more than two decades after Yunus disappeared, one witness had deposed in two years of actual proceedings.

Sachin Vaze, the primary accused in the Yunus murder, later became notorious in an entirely separate matter — the Antilia bomb scare case of 2021 involving industrialist Mukesh Ambani — and was eventually arrested by the NIA. His reinstatement to the force while facing a murder trial, and the protection offered to him and others by the state machinery over two decades, is one of the most systematic failures of prosecutorial independence in recent Maharashtra history.

"They killed Khwaja Yunus, burnt him, inflicted so much pain on him in the jail and even tried to erase this case," Asiya Begum said in an interview. "I will continue to fight for justice till the murderers of my son are punished." Her husband died waiting. She turned 80 still waiting.

Jayaraj and Bennix, Sathankulam, 2020

On the evening of June 19, 2020, P. Jayaraj and his son J. Bennix were taken from their mobile phone shop in Sathankulam, Thoothukudi district, Tamil Nadu. The stated reason was a violation of Covid-19 lockdown rules: the shop had allegedly been open a few minutes past the permitted curfew. The Central Bureau of Investigation would later establish, in its chargesheet, that this alleged violation had not even occurred.

What happened inside the Sathankulam police station over the next four days is now a matter of forensic record. Both men were subjected to prolonged and severe physical assault. Post-mortem reports documented eighteen serious injuries on each body. DNA analysis matched blood samples collected from the station's walls and floors to the victims. Witnesses testified that both men had been forced to clean their own blood from the floor. Blood-stained clothing was disposed of by the police. Jayaraj died on June 22. Bennix died the following day.

The police, as always, had a story. The story was that both men had fallen ill in custody and died of natural causes. The story was, as always, a lie.

Three cases. Three states. Three families. The same cover story, the same missing or destroyed evidence, the same grieving family at the courthouse door. In 1993, the lie was "he escaped." In 2002, the lie was "he escaped." In 2020, the lie was "they fell ill." The vocabulary shifts slightly. The structure of the lie — and the institutional machinery that sustains it — does not.

PART II

What the Law Said — and What It Could Not Do

The Sathankulam Verdict: A Constitutional Reckoning

On April 6, 2026, the First Additional District and Sessions Court in Madurai sentenced all nine convicted police personnel in the Sathankulam case to death. The judge, G. Muthukumaran, invoked the "rarest of rare" doctrine — the threshold established by the Supreme Court of India in Bachan Singh v. State of Punjab (1980) — which permits the imposition of capital punishment only in cases of such exceptional depravity that any lesser sentence would be inadequate to satisfy the collective conscience of society.

"If ordinary citizens had committed the same crime, ordinary punishment could have been given — but the police themselves have committed the crime." — Judge G. Muthukumaran, First Additional District and Sessions Court, Madurai, April 6, 2026

This sentence deserves careful attention. It is not merely rhetorical. It articulates a legal principle of considerable importance: that a crime committed by those vested with state power is of a categorically different character from the same crime committed by a private citizen. The betrayal of duty, the abuse of a position of trust, and the deliberate targeting of a person who is by definition defenceless — in a police lock-up, a person cannot resist, cannot call for help, cannot leave — all constitute aggravating factors that raise the moral gravity of the act.

Under Articles 21 and 22 of the Constitution of India, every person in custody retains the fundamental right to life and is protected against arbitrary detention. The police officer's uniform is not a private garment. It is a delegation of the State's monopoly on coercive force. When that delegation is used to inflict suffering and death, the crime is not only murder under the Bharatiya Nyaya Sanhita. It is a constitutional violation — the State killing the very person it was obligated, under its own founding document, to protect.

The Sathankulam verdict is, in this sense, historically significant. It is the first time in recent memory that an Indian court has applied the "rarest of rare" standard to a custodial death — treating the murder of persons in police custody not merely as murder, but as an act of exceptional constitutional gravity. That the Madurai sessions court arrived at death sentences for nine police officers is a verdict that the profession and the public should understand in its proper weight: it is not an exercise of vengeance. It is the law stating, with maximum force available to it, that this category of crime falls outside the range of ordinary punishment.

The Role of the High Courts: Conscience Where the State Failed

In both the Rajakannu case and the Sathankulam case, it was the Madras High Court — not the police, not the district administration, not the state government — that served as the institutional conscience of the system.

In Rajakannu's case, the High Court accepted a habeas corpus petition filed by an illiterate tribal woman, allowed the procedural innovation of witness examination, and ordered a Special Investigation when the police investigation proved compromised. In Sathankulam, the Madurai Bench of the Madras High Court took suo motu cognisance of the deaths on June 24, 2020 — unbidden, on the basis of news reports — expressed a formal loss of confidence in the state police's capacity to investigate itself, and directed that the case be transferred to the CBI. Without that intervention, there would almost certainly have been no trial.

In Maharashtra, the Bombay High Court's Aurangabad Bench performed the same function in the Somnath Suryawanshi case in 2025. Suryawanshi was a 35-year-old Dalit law student — a member of the Wadar community, listed as a Nomadic Tribe — who was arrested on December 12, 2024, during protests in Parbhani following the desecration of a replica of the Constitution near a statue of Dr. B.R. Ambedkar. He died in judicial custody within 72 hours of his arrest.

The Maharashtra government's initial account — that Suryawanshi had died of chest pain arising from a pre-existing respiratory condition — was directly contradicted by the post-mortem report, which established death by physical assault. His mother alleged that police officers had pressured her to conduct the last rites quietly, away from Parbhani, and had offered her Rs. 50 lakh to remain silent and not file a complaint. The High Court, in July 2025, ordered the registration of an FIR and directed investigation by an officer of the rank of Deputy Superintendent of Police.

The Maharashtra government's response was, by any measure, extraordinary. It challenged the High Court's FIR order before the Supreme Court of India — arguing, in open court, that the order was premised on a "presumption of guilt against the police authorities." The Supreme Court upheld the Bombay High Court. The FIR had still not been registered within the court's stipulated deadline at the time of writing.

This sequence — High Court orders an FIR; state government challenges the order in the Supreme Court; Supreme Court upholds the High Court; FIR still not registered — is not an aberration. It is the operating logic of custodial death accountability in India. Courts can order. States can resist. And the machinery of delay, available only to those with institutional power, grinds through.

D.K. Basu and the Paper Safeguards

In 1997, the Supreme Court of India delivered one of the most practically important judgments in the history of criminal procedure: D.K. Basu v. State of West Bengal. The Court, responding to a petition by a legal aid society, laid down eleven mandatory requirements to be followed in connection with every arrest and detention: a written memo of arrest signed by a witness; notification to a family member or friend; a medical examination of the arrestee at the time of arrest and at regular intervals thereafter; production before a magistrate within twenty-four hours; and the right to consult a lawyer, among others. The Court also held that any violation of these requirements would be treated as contempt of court, with the liberty to initiate proceedings against disobeying officers.

The D.K. Basu guidelines have now been in force for nearly three decades. In those three decades, no significant body of contempt proceedings against police officers for violation of these guidelines has developed in any High Court in India. Medical examinations are skipped. Production before magistrates is delayed. Families are not informed. The written arrest memo — perhaps the simplest requirement of all — is routinely absent.

"If you ask me 28 years ago and now, I don't see any big change in terms of the attitude of the police. Some of these circulars remain only on a paper proposition." — Justice K. Chandru, former Judge, Madras High Court; advocate in HCP No. 711 of 1993 (the Rajakannu custodial death case — the case behind Jai Bhim)

Justice Chandru's assessment is not the bitterness of a disappointed idealist. It is the considered judgment of a man who has seen this system from both sides of the Bar — as an advocate who forced a police cover-up open through the courts, and as a judge who spent years watching what happened when the same courts were not watching. When he says the circulars remain on paper, he means that the culture of enforcement has not changed, and that paper obligations without institutional consequences are not obligations at all.

PART III

A Verdict Is Not a Solution

What the Numbers Say

The Sathankulam verdict has been described, in the press and in public commentary, as a turning point. To understand why it is not — or at least, not yet — it is necessary to sit with the numbers that surround it.

170 custodial deaths in India in the first 74 days of 2026 alone. (Source: Ministry of Home Affairs, Lok Sabha, March 24, 2026) Maharashtra: 101 custodial deaths in 5 years — the highest of any state in India. Zero murder convictions in custodial death cases in four decades. (Source: NHRC data; Newslaundry / The News Minute, July 2025) Across India: NHRC recorded disciplinary action in exactly 1 custodial death case in the same 5-year period — in Tamil Nadu. (Source: Lok Sabha data, March 2026) Between 1999 and 2023: over 2,200 custodial deaths documented. No convictions in custodial death cases recorded between 2018 and 2023. (Source: NCRB)

These numbers tell a story that the Sathankulam verdict, as significant as it is, cannot by itself rewrite. The verdict is an extraordinary exception in a pattern of near-total impunity. It arrived because the Madras High Court intervened suo motu. Because the case was transferred to the CBI. Because the forensic evidence survived the police's attempts at destruction. Because the national outrage following the deaths created political pressure that kept the prosecution on track. Remove any one of these factors, and the case follows the trajectory of the other two thousand.

Maharashtra is the most instructive case study in what the normal trajectory looks like. In the Khwaja Yunus case: the government sanctioned prosecution of four out of fourteen named accused. The special prosecutor who sought to expand the accused was removed. The accused were reinstated while the trial was pending. Over twenty-three years, the trial has produced one witness deposition. No conviction. No closure. Asiya Begum is past eighty.

In the Somnath Suryawanshi case, a state government went to the Supreme Court of India to prevent an FIR from being registered against its own police. That is not a failure of the law. That is the State actively using the law — appeals, procedural challenges, institutional friction — as a weapon against accountability.

The Structural Gaps That No Verdict Can Fill

Any honest legal assessment of custodial deaths in India must identify the structural conditions that produce them, because those conditions will produce more deaths regardless of what happens to the nine officers convicted in Madurai.

The first and most fundamental gap is the absence of a standalone anti-torture statute. India has no law that defines torture, criminalises it as a specific and aggravated category of offence, and prescribes enhanced punishment that reflects the unique wrong involved when a state agent inflicts deliberate suffering on a person in custody. Custodial deaths are prosecuted under the ordinary murder provisions of the Bharatiya Nyaya Sanhita — the same provisions applicable to a brawl in a street. There is no recognition in the criminal law of the constitutional dimension of the wrong, the power asymmetry involved, or the institutional nature of the failure. A law that treats the murder of a person in a police lock-up identically to the murder of a person in a road rage incident is not merely inadequate. It is, in an important sense, dishonest about what custodial killing actually is.

The second gap is India's longstanding refusal to ratify the United Nations Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment. India signed the Convention in 1997. As of April 2026, it has not ratified it — making India one of a small number of countries that has signed but not ratified one of the core UN human rights instruments. Ratification would not be merely symbolic. It would bind India under international law to maintain domestic legislation criminalising torture, to investigate all credible allegations, to ensure non-refoulement, and to submit to periodic review by the UN Committee Against Torture. Twenty-nine years of non-ratification is not an oversight. It is a choice.

The third gap is structural: in India, the police investigate themselves. When a person dies in a police station, the first investigation is almost always conducted by the police — the same institution whose officers are implicated. The mechanisms for overriding this — CBI transfer, judicial inquiry under Section 176 BNSS, Special Investigation Teams — require active judicial intervention to trigger. They are not the default. They are the exception, available only when a case attracts enough attention to force a court's hand. The overwhelming majority of custodial deaths attract no such attention.

The fourth gap is the prosecution sanction requirement under Section 218 of the Bharatiya Nagarik Suraksha Sanhita (which replaced Section 197 of the Code of Criminal Procedure). Under this provision, no court can take cognisance of an offence alleged to have been committed by a public servant in the discharge of official duties without the prior sanction of the appropriate government. This provision was designed to protect honest public officials from vexatious litigation. In the context of custodial deaths, it has been systematically weaponised to protect dishonest ones. The government selectively grants sanction — as in the Yunus case, where fourteen accused were named and four were sanctioned — and the courts have been reluctant to interfere with this executive discretion.

The fifth gap is the magistracy. Under the D.K. Basu guidelines and the provisions of the BNSS, a judicial magistrate is the first independent authority who encounters an arrested person — within twenty-four hours of arrest. The magistrate's role at that encounter is not ceremonial. It includes ensuring that the arrestee has been medically examined, that family has been notified, that the arrest is properly documented, and that the person has access to legal counsel. In practice, production before magistrates is frequently delayed, the inquiry is cursory, and the magistrates — who are part of the same state executive apparatus that employs the police — often lack both the independence and the institutional training to perform this function effectively.

What Reform Must Look Like

The legal profession has been identifying these gaps since at least the National Police Commission's recommendations in 1979 and 1981. The Prevention of Torture Bill was introduced in Parliament in 2010, examined by a Rajya Sabha Select Committee, and then quietly allowed to lapse. It was never revived.

What is required is not novel. It is available in the legislative and institutional experience of comparable democracies, and it is consistent with India's own constitutional commitments. A standalone anti-torture statute must define custodial torture and custodial death as distinct and aggravated offences, must vest the power to prosecute in an authority independent of the state government implicated, and must provide for mandatory minimum sentences that reflect the constitutional dimension of the wrong. The UN Convention Against Torture must be ratified — not in the next decade, but now, and the domestic legislative changes necessary for ratification must accompany it. An independent statutory authority for the investigation of all custodial deaths — insulated by design from the police chain of command — must be established. The prosecution sanction requirement under Section 218 BNSS must be reformed so that it cannot serve as a veto over prosecution in cases where the allegation is custodial torture or death. And the magistracy must be trained, resourced, and institutionally protected so that the twenty-four-hour production requirement functions as the constitutional safeguard it was designed to be.

None of this will happen quickly. Institutional reform of the police, the magistracy, and the prosecution system in India is among the most politically and administratively complex tasks imaginable. But the alternative — continuing to generate landmark verdicts that arrive, on average, once every thirty years, for crimes that occur every day — is not a legal system. It is a monument to selective accountability.

CLOSING

The Question the Law Must Answer

The Sathankulam death sentences will travel to the Madras High Court for mandatory confirmation. They will almost certainly reach the Supreme Court on appeal. Whatever the final quantum of punishment, the verdict has already accomplished something permanent: it has placed on the legal record of this country that nine uniformed officers of the State of Tamil Nadu murdered two men entrusted to their care, and that a court of law found this to be among the worst crimes a human being can commit.

That is meaningful. The Rajakannu case established, in 1994, that the police cannot make a man disappear and then claim he escaped. The Sathankulam case has now established that the police cannot torture a father and son to death in a lock-up and walk away. These are not small things. They are points of legal reference that change what is possible in Indian courts.

But Justice Chandru, who has spent more time thinking about this question than almost anyone alive, said what needs to be said when he reflected on the Sathankulam verdict: twenty-eight years after Rajakannu, he does not see a big change in the attitude of the police. The circulars remain on paper.

Parvathi fought for thirteen years. Asiya Begum has been fighting for twenty-three. The family of Somnath Suryawanshi is fighting now. These women — poor, marginalised, without institutional power — have carried the burden of building custodial accountability in India almost entirely on their own shoulders, through courts that were designed for the powerful and doors that were never meant to open for them.

A verdict punishes the past. Only structural reform prevents the future.

India has the law. It has the courts. It has the constitutional framework. What it does not have is the institutional will to make the law mean the same thing for the person in the lock-up as it does for the person who put them there.

The script never changes. The question — the only question that matters — is whether the law will.

REFERENCES & LEGAL CITATIONS

Cases

1. Bachan Singh v. State of Punjab, (1980) 2 SCC 684 — Supreme Court of India ("rarest of rare" doctrine for capital punishment)

2. D.K. Basu v. State of West Bengal, (1997) 1 SCC 416 — Supreme Court of India (mandatory arrest safeguards; contempt jurisdiction for violations)

3. HCP No. 711 of 1993 — Madras High Court (Parvathi v. State of Tamil Nadu; Rajakannu custodial death; Advocate K. Chandru for petitioner; Justices P.S. Mishra and Shivaraj V. Patil)

4. Sessions Court, Madurai — Judgment dated April 6, 2026 (Sathankulam custodial death; Judge G. Muthukumaran; nine police officers sentenced to death)

5. Vijayabai Vyankat Suryawanshi v. State of Maharashtra, 2025 SCC OnLine Bom 2588 — Bombay High Court, Aurangabad Bench, July 4, 2025 (Somnath Suryawanshi custodial death; FIR directed)

6. State of Maharashtra v. Vijayabai Vyankat Suryawanshi — Supreme Court of India, July 30, 2025 (upholding Bombay HC order; state challenge dismissed)

Legislation

7. Articles 21 and 22, Constitution of India (right to life and personal liberty; protection against arbitrary arrest and detention)

8. Section 218, Bharatiya Nagarik Suraksha Sanhita, 2023 (prosecution sanction for public servants; replaces Section 197, Code of Criminal Procedure, 1973)

9. Section 176, Bharatiya Nagarik Suraksha Sanhita, 2023 (mandatory judicial inquiry into custodial deaths)

10. United Nations Convention Against Torture and Other Cruel, Inhuman or Degrading Treatment or Punishment, 1984 (signed by India, 1997; not yet ratified as of April 2026)

Reports & Data

11. Ministry of Home Affairs, Government of India — response to Lok Sabha Unstarred Question, March 24, 2026 (state-wise custodial death data, NHRC figures 2021-22 to 2025-26)

12. National Crime Records Bureau (NCRB) — Prison Statistics India; custodial death data 1999-2023

13. National Human Rights Commission of India — annual reports; custodial death data by state

14. Newslaundry / The News Minute — "4 Decades, Hundreds of Custody Deaths, No Murder Conviction: The Curious Case of Maharashtra," July 23-24, 2025

15. National Herald India — "Maharashtra Records Highest Number of Custodial Deaths, Raising Renewed Concerns," July 2026

16. Global Torture Index 2025: India Factsheet — Association for the Prevention of Torture / World Organisation Against Torture (OMCT)

17. National Police Commission Reports, 1979-1981 (First through Eighth Reports; recommendations on police reform and custodial accountability)

18. Prevention of Torture Bill, 2010 — introduced in Lok Sabha; referred to Rajya Sabha Select Committee; lapsed without enactment

Comments