The 2015 and 2018 amendments turned a shield into a sword. Police data from Rajasthan and Haryana now record close to half of atrocity FIRs as false, convictions hover around a third, and the Supreme Court has just declined to look again.
New Delhi: On September 22, a Bench of Chief Justice of India Surya Kant, Justice Joymalya Bagchi and Justice V. Mohana declined to entertain a petition challenging the 2018 amendment to the Scheduled Castes and Scheduled Tribes (Prevention of Atrocities) Act, 1989. The amendment had restored the bar on anticipatory bail and removed the requirement of a preliminary inquiry before an FIR is registered under the Act. The Court said the amendment had already been upheld in 2020, that sufficient safeguards exist because courts can examine whether a prima facie case is made out, and that it did not accept the argument that the Act was producing reverse discrimination.
That is a procedurally correct answer. It is also an incomplete one. The petition may have been badly framed, and a challenge that has already failed before a three-judge Bench cannot be relitigated by a fresh writ. But the underlying problem the petitioner was pointing at is real, it is measurable, and it has grown worse in the eight years since Parliament overrode the Supreme Court’s own attempt to fix it. This piece is an argument for why the institution should find another route back to the question.
How the Act became what it is
The 1989 Act was Rajiv Gandhi’s answer to a genuine failure of the ordinary criminal law: caste violence in rural India was going unreported, unprosecuted and unpunished. It created special offences, special courts and, in Section 18, a bar on anticipatory bail under Section 438 of the CrPC. For a quarter of a century that bar was read by the Supreme Court in State of M.P. v. Ram Krishna Balothia (1995) as constitutionally valid, on the reasoning that persons accused of caste atrocities form a distinct class.
The Modi government then amended the law twice. The 2015 amendment, in force from January 2016, was the larger rewrite. It added new offences (garlanding with footwear, forcing manual scavenging, imposing social boycott, obstructing access to public places and resources), created exclusive special courts and exclusive public prosecutors, inserted Section 8’s presumption that an accused who knew the victim’s family is presumed to have acted on caste motive unless he proves otherwise, and, through Chapter IVA, gave victims and witnesses a statutory bundle of rights that no other criminal statute confers.
The 2018 amendment was smaller in text and larger in consequence. It was Parliament’s direct response to Dr Subhash Kashinath Mahajan v. State of Maharashtra (March 20, 2018), in which Justice A.K. Goel and Justice U.U. Lalit had laid down three procedural filters: a preliminary inquiry within seven days before an FIR under the Act, prior approval of the appointing authority before arresting a public servant and of the Senior Superintendent of Police before arresting anyone else, and anticipatory bail where no prima facie case existed. The judgment provoked the April 2, 2018 Bharat Bandh, in which at least nine people died, most of them in the Gwalior-Chambal belt, and within four months Parliament had inserted Section 18A. It says, in terms, that no preliminary inquiry is required, that no approval is needed for arrest, and that Section 438 shall not apply “notwithstanding any judgment or order or direction of any Court”. In October 2019 a three-judge Bench recalled the Mahajan directions on review, and in Prathvi Raj Chauhan v. Union of India (February 2020) another three-judge Bench upheld Section 18A, with the sole comfort that a court retains its inherent power to quash an FIR that discloses no offence on its face.
So the position today is this: an allegation that someone insulted a Scheduled Caste person “within public view” produces an FIR without inquiry, an arrest without supervisory approval, and no anticipatory bail. The person then finds out whether the allegation is true from inside the process. That is what the Chief Justice’s Bench described as adequate safeguards.
What the police records say about the cases
The Act’s defenders often say that misuse is an upper-caste myth. The data belongs to the police, not to either side, and it is worth setting out.
The NCRB’s Crime in India series shows registrations under the Act climbing from 45,995 in 2020 to 52,866 in 2022 and a peak of 53,372 in 2023, before easing to 48,669 in 2024. Convictions, measured on cases where trial concluded, have hovered around a third: 34.1 per cent in 2022, 32.4 per cent in 2023 and 33.9 per cent in 2024. Against the overall IPC conviction rate, which has run at 55 to 57 per cent in recent years, that is a large gap. And the trial backlog under the Act has crossed 2.6 lakh cases.
The more telling numbers are the police “final reports”, the closures recorded before a case ever reaches a courtroom. The Ministry of Social Justice’s own annual report under the Act for 2022 records that 14.78 per cent of SC-related cases and 14.71 per cent of ST-related cases were closed in final reports, a category that expressly includes “false” and “mistake of fact”. Charge sheets were filed in only about 60 per cent.
At the State level the picture is starker. The NCRB’s 2023 data, as analysed by The Tribune, shows the Rajasthan Police declaring 49.3 per cent of the atrocity FIRs it disposed of as false, and Haryana second at 45.9 per cent. In Haryana, 702 of 1,529 SC/ST Act FIRs disposed of that year were closed as false; of the criminal intimidation cases clubbed with the Act, 62.3 per cent were closed as false. Haryana’s State Crime Records Bureau had already put the false-case figure for atrocities at 50.6 per cent in 2021. In 2020, the Rajasthan Police itself publicly stated that over 40 per cent of the SC/ST Act cases registered that year had been found fake on investigation.
Two caveats are necessary, and the Act’s defenders are entitled to them. A police “false” closure is a police conclusion, not a judicial finding, and it is made by a force that Dalit organisations say is itself caste-biased and reluctant to investigate. And these are the same States that lead the country in false closures across all crime categories. But the numbers are what the State’s own machinery is recording. If half the FIRs a State registers under a statute are being closed by that State’s police as false, the statute has a design problem, not merely an implementation problem. And under Section 18A, every one of those complainants had the power to trigger an arrest without anyone checking first.
The constitutional objection
The argument is not that the Act should go. It is that Section 18A, read with Section 18, offends the basic structure of criminal procedure in a way the Supreme Court has recognised in every context except this one.
Set out plainly, the provisions of the Constitution that the Act, as amended, comes into conflict with are these. Article 14, the guarantee of equality before the law, because Section 18 denies anticipatory bail to an entire class of accused defined by the caste of the complainant rather than by the gravity of the offence, and because Section 18A treats a petty insult and a caste massacre identically at the stage of arrest. Article 19(1)(a), the freedom of speech and expression, because Sections 3(1)(r) and 3(1)(s) criminalise words alone, and Section 18A ensures that the mere allegation of those words leads to arrest before anyone tests whether they were said, or whether they were said because of caste. Article 21, the guarantee that no person shall be deprived of liberty except by a procedure that is fair, just and reasonable, because Section 18A(1)(b) removes the arrest safeguards that Arnesh Kumar read into that Article, and because Section 8 reverses the presumption of innocence, which the Supreme Court in Noor Aga v. State of Punjab (2008) recognised as a facet of Article 21. Article 22, the constitutional protections on arrest and detention, because a bar on anticipatory bail combined with a bar on preliminary inquiry leaves the accused with no forum before custody. And the Article 32 and Article 226 remedy itself, because the only relief the law leaves open, a High Court quashing petition, is one that operates after the harm and is available in practice only to those who can afford it. Against all this stands the Act’s own constitutional anchor, Article 17, which abolishes untouchability and makes its practice an offence, read with Article 15(4) and Article 46, which mandate special provision for the Scheduled Castes and Tribes. Nobody disputes that anchor. The question is whether Article 17 licenses a procedure that Articles 14, 19 and 21 would not permit for any other offence in the statute book.
Start with Article 21. In Arnesh Kumar v. State of Bihar (2014) the Supreme Court held that arrest is not to be automatic on registration of a cognisable offence, that a police officer must record reasons that satisfy the necessity tests in Section 41 of the CrPC, and that a magistrate must be satisfied before authorising detention. Those directions were issued in a Section 498A case precisely because the Court found that provision being used as a weapon. Section 18A(1)(b) legislates that no approval is required for arrest under the Act. Whether Arnesh Kumar‘s necessity test survives that text at all is unsettled; in practice, on the ground, it does not.
Then Article 14. The bar on anticipatory bail treats a man accused of abusing his neighbour in a property dispute, where the neighbour happens to be a Scheduled Caste, identically with a man accused of murder, arson or gang rape motivated by caste. There is no gradation. Balothia justified the classification in 1995 on the footing that atrocities are a distinct class of crime. It said nothing about a class of accused defined solely by the caste of the complainant, which is what Section 18 produces in practice. The 2018 amendment’s “notwithstanding any judgment” clause is also, on its face, a legislative override of a specific judicial direction rather than a change to the legal basis of that direction, which is the line the Supreme Court has drawn since Madan Mohan Pathak v. Union of India (1978).
Then the presumption of innocence, which the Supreme Court treats as part of Article 21. Section 8 presumes caste motive wherever the accused had personal knowledge of the victim or her family. In a village, in a workplace, in a university department, everyone knows everyone. The presumption is thus close to universal, and the burden is on the accused to disprove a state of mind.
The Chief Justice’s Bench pointed to the court’s power to test whether a prima facie case exists. That is the Prathvi Raj Chauhan safeguard. It operates, however, after arrest, after custody, after the FIR has been reported in the local press with the accused’s name and caste, and it requires the accused to fund a Section 482 (now Section 528 BNSS) petition in the High Court. The Supreme Court’s own judgments show how often that petition succeeds: in Hitesh Verma v. State of Uttarakhand (2020) the Court quashed a case where the alleged insult occurred inside a private house; in Ramawatar v. State of M.P. (2021) it quashed one arising out of a village property quarrel; in Shajan Skaria v. State of Kerala (2024) it held that a Section 3(1)(r) charge requires the insult to be because of caste, not merely directed at a person who is of a Scheduled Caste; and as recently as January 2025 it quashed proceedings where the FIR, read as a whole, disclosed nothing about caste at all. Each of those cases is a person who was arrested, or lived under threat of arrest, for months or years before the Court said no offence had ever been disclosed. A safeguard that arrives after the harm is not a safeguard against the harm.
The human cost and the social cost
The instances are not hard to find, and they cut both ways, which is the point.
In June 2026, a 41-year-old fruit vendor in Bengaluru died after consuming poison, leaving a video naming persons he said had been extorting money from him under threat of an atrocity case; the police registered abetment charges against them. In Jaipur in 2022, a young chartered accountant died by suicide leaving a note that his family had been “treated like criminals” after a case under the Act was registered against his mother in a money dispute. In Churu the same year, a businessman died after, according to the FIR, a group threatened him with rape and atrocity complaints unless he paid up. And this month, at IIT Bombay, Professor Suryanarayana Doolla, a decorated teacher, was booked under the Act and for abetment of suicide after a second-year student who had been caught cheating in an examination died by suicide; the student’s parents allege months of caste abuse, the faculty forum says the student never approached the institute’s SC/ST cell, and an early police probe has reportedly found no concrete evidence against him. The allegations remain under investigation, and it may yet turn out that the parents are right. But the professor has already been removed as Dean and sent on leave on the strength of an FIR that the law forbade anyone from testing before it was registered.
The Act has also cost lives on the other side of the line, and honesty requires saying so. On April 2, 2018, Rahul Pathak, a student standing on the balcony of his own house in Morena, was killed by bullets allegedly fired by demonstrators protesting the Mahajan judgment. In Shabbirpur, Saharanpur, in May 2017, a Thakur youth was killed and Dalit homes were burnt in clashes that began over a procession. At Bhima Koregaon on January 1, 2018, a Maratha youth died in caste violence that spread across Maharashtra. None of these deaths was caused by a statute. But each was the product of a politics in which caste has become the currency of grievance, and in which a criminal law is experienced by one community as its only protection and by another as an instrument of hostage-taking. When a law makes the mere allegation decisive, it hands everyone with a grievance a weapon, and it is not surprising that the weapon is used in land disputes, in office rivalries, in matrimonial fights and in campus politics. That is what a cold war looks like: not open violence, but a settled mutual suspicion in which every interaction across the caste line carries a legal risk that runs in only one direction.
What the Court could still do
The Chief Justice was right that the Court cannot strike down what a three-judge Bench upheld in 2020 at the instance of a fresh petitioner. But Prathvi Raj Chauhan was itself a narrow decision, and Justice Ravindra Bhat’s concurring opinion in it expressly recorded that anticipatory bail could be granted where the complaint, on its face, did not make out an offence. What has never been done is the reconciliation of Section 18A with Arnesh Kumar, the question whether Section 8’s presumption is proportionate after K.S. Puttaswamy and Anuradha Bhasin, and whether a mandatory FIR-without-inquiry regime is constitutional for the lower-tier offences in Section 3(1) while being defensible for the grave ones. A reference on those questions is open. So is a direction, on the administrative side, that States publish their false-closure data under the Act district by district, so that the debate stops being one of anecdote against anecdote.
Parliament, too, has room. The Act contains no offence of filing a false atrocity complaint; the ordinary Section 217 BNS prosecution for false information is almost never launched. A provision with teeth, restricted to complaints the special court itself finds fabricated, would deter the extortion racket without touching a single genuine victim.
The other side, stated fairly
Those who defend Section 18A as it stands make three points, and they deserve to be met rather than dismissed. First, the low conviction rate, they say, reflects hostile witnesses, caste-sympathetic investigating officers and prosecutors who do not turn up, not false complaints; the Wire’s analysis of the 2016 data made this case, and the Ministry’s own reports concede that only 194 of 498 districts in 14 States have functioning exclusive special courts. Second, the 2018 Bharat Bandh was not manufactured outrage; it was the reaction of communities who read Mahajan as a message that their word would be doubted before anyone else’s. Third, the “false” closures are recorded by police forces that Dalit organisations have documented, for decades, as refusing to register genuine complaints in the first place, so a high false-closure rate may measure police attitude as much as complainant conduct. Each of these is a serious argument. None of them answers why an arrest should be made before anyone checks whether the words alleged were even said.
The Supreme Court closed a door on September 22. It should be conscious that it has not answered the question standing behind it.
Views are the author’s own.