Section 18A of the Scheduled Castes and the Scheduled Tribes (Prevention of Atrocities) Act, 1989 exists because of an argument about who writes criminal procedure. In March 2018 a two-judge bench of the Supreme Court, persuaded that innocent people were being arrested on unexamined accusations, read three safeguards into a statute that contains none of them: a screening enquiry before a first information report, an approval before arrest, and a softened bar on anticipatory bail. Within five months Parliament had removed all three by an amendment drafted in terms that expressly displaced the judgment. Within twenty-three months the Supreme Court had recalled its own directions as an act of judicial legislation and then upheld the amendment that had displaced them. No other episode in modern Indian law runs the full circuit of interpretation, override, repentance and validation so quickly, or leaves the boundary between a court’s power to interpret and a legislature’s power to enact quite so clearly marked.

The question the whole sequence turns on is deceptively simple. When a protective statute is being misused, and the legislature has chosen not to write any filter into it, may a court write the filter itself? The March 2018 bench answered yes, reasoning from the guarantee of a just and fair procedure under Article 21 of the Constitution. The review bench answered no, reasoning that a court which adds conditions to a statute has stopped interpreting it and started enacting it. Between those two answers lies almost everything a reader needs to understand about how Indian courts and Parliament divide their work, and the practical stakes are not abstract. The question that brings most readers to this subject, whether a person accused under the Act can obtain anticipatory bail, has been answered wrongly in both directions by a large volume of published commentary: it is not freely available, and it is not barred in every case. Stating the position exactly, and showing how each of the four steps in the sequence contributed to it, is the work of this article.

The statute the Court was reading

The Scheduled Castes and the Scheduled Tribes (Prevention of Atrocities) Act, 1989, Act 33 of 1989, brought into force on 30 January 1990, is a protective statute of a deliberate and unusual design. It does not merely punish caste-based offences through the ordinary criminal process. It creates its own catalogue of offences in Section 3, its own special courts, its own machinery of relief and rehabilitation, and its own procedural choices, each of which was made by Parliament in the knowledge that the ordinary process had failed the communities the Act protects. A full account of that design, of what the Act makes an offence and of the rights it gives victims and witnesses, belongs to the statute pillar on the SC and ST Atrocities Act 1989, and this article takes that account as its foundation rather than repeating it. The wider field within which the Act sits is mapped in the guide to social justice legislation in India.

One procedural choice matters here above all others. Section 18 of the Act, in its original form, provided in full: “Nothing in section 438 of the Code shall apply in relation to any case involving the arrest of any person on an accusation of having committed an offence under this Act.” Section 438 of the Code of Criminal Procedure, 1973 is the provision under which a person who apprehends arrest may ask a court of session or a High Court to direct that, in the event of arrest, they be released on bail. It is commonly called anticipatory bail. Parliament’s choice in Section 18 was therefore not an oversight or a gap. It was a decision, taken in the text of the statute, that the ordinary pre-arrest protection available in every other prosecution would not be available in a prosecution under this Act, because the legislature judged that the prospect of immediate arrest was part of what gave the protective scheme its force, and that a complainant from a Scheduled Caste or a Scheduled Tribe, complaining against a socially stronger accused, needed the process to begin without a preliminary contest about bail.

That choice was challenged and upheld. In State of Madhya Pradesh v. Ram Kishna Balothia, decided in 1995 and reported at (1995) 3 SCC 221, the Supreme Court sustained Section 18 against the argument that excluding anticipatory bail violated the equality and liberty guarantees, accepting the legislature’s premise that the exclusion served the Act’s protective object. The position that then stood for over two decades was accordingly plain on the face of the statute. There was no enquiry stage before a first information report beyond what the general law required, no official whose approval stood between an accusation and an arrest, and no anticipatory bail, however weak the accusation might be said to be. An accused person’s remedies were the ordinary ones that operate after the process begins: bail after arrest, discharge, and the power of the High Court to quash a proceeding that discloses no offence.

The statute was also not static. The Amendment Act of 2015, Act 1 of 2016, brought into force on 26 January 2016, substantially expanded the catalogue of offences in Section 3, added provisions on the rights of victims and witnesses, and restated much of the machinery. It is against that restated Act, barely two years old, that the judicial sequence of 2018 must be understood. Parliament had revisited the statute comprehensively in the revision brought into force in January 2016, and had again chosen not to write any screening mechanism, any approval requirement, or any anticipatory bail route into it. That considered silence is the single most important fact in the interpretive dispute that followed, because the two sides read it in opposite ways. For the bench that acted in March 2018, the silence was a gap that produced injustice in individual cases and that the Court could fill. For the bench that recalled the directions in October 2019, the silence was a legislative decision that the Court had no authority to reverse by direction.

To understand what the exclusion of that provision costs an accused person, it helps to know why the provision exists. Anticipatory bail was recommended by the Law Commission in its Forty-first Report and was introduced into the Code of Criminal Procedure, 1973 as a deliberate innovation, on the reasoning that influential persons were sometimes able to set the criminal process in motion against their rivals on frivolous accusations, and that a person so targeted should not have to undergo arrest and custody before a court could examine the accusation. The Constitution Bench in Gurbaksh Singh Sibbia v. State of Punjab, (1980) 2 SCC 565, treated the provision as a safeguard of personal liberty to be applied according to its terms and the facts of each case, and declined to confine it by judicially invented conditions. The symmetry with 2018 is instructive and seldom noticed. In Sibbia the Court refused to write conditions into a liberty-protecting provision that Parliament had enacted in general terms. In Mahajan the Court wrote conditions into a liberty-restricting provision that Parliament had enacted in absolute terms. The review bench’s jurisprudence treats the two moves as mirror images of the same error: the judicial task is to apply the provision the legislature wrote, at the width the legislature chose, and the remedy for a width that produces injustice in a class of case lies with the institution that chose the width. That is a coherent account of the judicial role, and it has the additional virtue of explaining why Sibbia and the recall sit comfortably together in the law reports although their practical valences are opposite.

What did the text of the Act leave open, if anything?

Very little. The Act fixed the procedure for a case under it and excluded Section 438 of the Code by name. What it did not address was the quality of accusations, the discipline of arrest, or a public servant accused for an official act, and it was into those silences that the 2018 bench directed its reasoning.

The distinction matters for everything that follows. A court interpreting an ambiguous provision performs the ordinary judicial task, and its reading binds because the text bore the meaning. A court supplementing a complete text with conditions the text omits performs a different task, and the legitimacy of that task depends on doctrines about gap filling, about the power under Article 142 to do complete justice, and about where interpretation ends and legislation begins. The 2018 sequence is the rare case in which the Supreme Court confronted that distinction in its own work twice within twenty months, first by crossing the line while describing the crossing as interpretation, and then by saying so in terms. The reader who keeps the distinction in view will find that the individual steps, which are often reported as a bewildering series of reversals, fall into a single intelligible argument.

Reading in, reading down, and legislating from the bench

Indian courts supplement statutory texts constantly, and most supplementation is uncontroversial because it takes forms the legal system has learned to recognise and to bound. A court reads down a provision when it chooses, between two available constructions, the one that keeps the provision within the Constitution, narrowing the words rather than striking them; the technique presupposes that the narrower meaning is one the enacted words can bear. A court reads in words when a provision read word for word would be unconstitutional or absurd, and a small insertion, consistent with the provision’s scheme, saves it; the technique is defended, where it is defended at all, as the completion of an evident legislative purpose. A court fills a gap when the statute is silent on a point its own operation requires to be answered, as with the procedural detail of a remedy the statute creates but does not regulate. And a court issues guidelines in a vacuum, on the Vishaka model, when there is no law at all on a subject the Constitution requires to be governed, and it does so expressly until the legislature acts, borrowing its authority from the constitutional rights engaged rather than from any statute.

The Mahajan directions fit none of these four forms comfortably, and the attempt to make them fit is the intellectual history of the episode. They were not a reading down of Section 18, because the bench did not hold that the exclusion of anticipatory bail bore a narrower meaning than its words; it held that the exclusion yielded where the complaint was wanting, which is a condition on the provision’s operation, not a construction of its language. They were not the filling of a gap, because the points they regulated, the stages before registration and before arrest, were already regulated by the Code of Criminal Procedure, whose scheme the directions displaced for this class of case. They were not vacuum guidelines, because the field was occupied by a comprehensive statute that had been revised in expanded form in 2016. They most resembled the fourth form in their tone, since the bench evidently regarded the territory of pre-arrest protection as one in which the Constitution had something to say that the statute had failed to hear, but the resemblance fails at the decisive point: Vishaka guidelines govern until legislation occupies the field, and here the field had been occupied for twenty-nine years, and reoccupied, in expanded form, two years before.

The review bench’s judgment is best understood as a sorting of these categories. It did not retreat from reading down, or from the arrest jurisprudence built on Article 21, or even from guidelines in a genuine vacuum, each of which it treated as established. It held that what had occurred in Mahajan belonged to none of them: conditions had been added to a complete scheme, in the teeth of the scheme’s own choices, on the strength of a disagreement with the legislature’s assessment of how the scheme worked in practice. That, the bench said, is legislation, and it is not open to a court however well founded the disagreement. The clarity of that sorting is why the October 2019 judgment, rather than the more dramatic amendment that preceded it, is the document a student of the episode should read first.

A sanction refusal that became a safeguards judgment

The case that produced the directions was not a challenge to Section 18, and it did not begin as a vehicle for reforming the Act. It began, as the judgment itself records, in an office dispute in Maharashtra. The appellant, Dr Subhash Kashinath Mahajan, was serving as Director of Technical Education in the State of Maharashtra. A subordinate employee of the department, who had earlier worked as a store keeper at the Government College of Pharmacy at Karad, had two senior officers who were not members of a Scheduled Caste, and those officers had made an adverse entry in his annual confidential report recording that his integrity and character were not good. On the strength of that entry the employee lodged a first information report against the two officers under the Act at the beginning of January 2006. Because the officers were public servants, the investigating officer applied to the Director of Technical Education for sanction to prosecute them under Section 197 of the Code of Criminal Procedure, and the appellant refused that sanction on 20 January 2011.

The employee’s response was to lodge a first information report against the appellant himself. His case was that the two officers were Class I officers for whom only the State Government could grant or refuse sanction, so that the Director, in deciding the question himself, had committed offences under the Act, the provisions invoked being those that then penalised giving false information and a public servant’s wilful neglect and related defaults, together with offences under the Penal Code. The appellant sought to have the proceedings quashed. The High Court refused. Before the Supreme Court, the case was reframed at a level far above its facts. Notice was issued to the Attorney General because the interpretation of a central statute was involved, and a senior counsel, Amrendra Sharan, was appointed to assist the Court as amicus curiae. The questions the bench posed for itself were whether a unilateral allegation could set the criminal process in motion against an officer who had acted in an official capacity, whether that was a just and fair procedure within Article 21, and whether procedural safeguards could be devised so that the Act’s provisions were not abused for extraneous considerations.

On those questions the appeal was, in one sense, easy. The Court held that the proceedings against the appellant were a clear abuse of the process of the court and quashed them. A decision on a sanction application, taken in the discharge of an official function, had been converted into an accusation of atrocity because the decision went against the complainant, and the Court was satisfied that the accusation could not be sustained. If the judgment had stopped there, it would have been an unremarkable exercise of the quashing power, noted by practitioners and forgotten by everyone else. It did not stop there. Drawing on material placed before it about the working of the Act, including figures of the National Crime Records Bureau which the bench read as showing that a substantial proportion of cases under the Act were closed after investigation as false and that convictions formed a small proportion of disposals, the bench concluded that the case disclosed a systemic problem: the acknowledged abuse of the law of arrest in cases under the Act. It answered that problem not with observations but with directions, issued to govern every case under the Act from that day forward, and expressed to operate prospectively.

The step from a quashed prosecution to a general procedural code for a statute is the step this entire episode is about, and it is worth pausing on how natural it can look from inside a courtroom. A bench that has just seen an apparently abusive prosecution, and that is shown figures suggesting the instance is not isolated, faces a genuine institutional temptation. It can decide the case and leave the pattern untouched, trusting a legislature that has shown no inclination to act. Or it can use the authority of the case to set the pattern right, in the belief that the Constitution’s liberty guarantee requires a filter and that nobody else will build one. The guidelines jurisprudence of the Indian Supreme Court, of which the workplace harassment directions in Vishaka are the celebrated example, supplies a ready vocabulary for the second course. What the March 2018 bench did not reckon with, and what the review bench later held against it, is that Vishaka filled a vacuum: there was no statute at all. Here there was a statute, re-enacted in expanded form in 2016, and the vacuum was not in the law but in the bench’s assessment of how the law worked.

Why did an appeal about one officer produce general directions?

Because the bench framed the questions at the level of the statute rather than the dispute. Once the issue was whether the Act’s procedures were a just and fair procedure under Article 21 in the run of cases, the answer could not be confined to the parties, and the figures were treated as evidence of a structural defect.

The directions of 20 March 2018

The judgment in Dr Subhash Kashinath Mahajan v. State of Maharashtra, reported at (2018) 6 SCC 454, was delivered on 20 March 2018 by a bench of Adarsh Kumar Goel and Uday Umesh Lalit JJ. Its operative conclusions, set out at the end of a long judgment that also directed governments to enforce the Act strictly, were five in substance. The proceedings in the case itself were quashed as an abuse of process. It was then held that there is no absolute bar against the grant of anticipatory bail in cases under the Act if no prima facie case is made out, or where on judicial scrutiny the complaint is found to be prima facie mala fide, and the bench approved the approach of the Gujarat High Court decisions that had taken that view and clarified the earlier judgments of the Court, including Balothia, which had been read as laying down an unqualified bar. Third, in view of what the judgment called the acknowledged abuse of the law of arrest in cases under the Act, the arrest of a public servant could be made only after the approval of the appointing authority, and the arrest of any other person only after approval by the Senior Superintendent of Police; the approval was to be granted in appropriate cases for reasons recorded, and those reasons were to be scrutinised by the magistrate before further detention was permitted. Fourth, to avoid the false implication of an innocent person, a preliminary enquiry could be conducted by the Deputy Superintendent of Police concerned to find out whether the allegations made out a case under the Act and were not frivolous or motivated. Fifth, a violation of the approval and enquiry directions would be actionable by way of disciplinary proceedings as well as contempt.

Each direction answered a real anxiety, and the judgment’s reasoning deserves to be stated in its strongest form before it is criticised, because the stance of this series is that a position is only worth overturning in the form its defenders would recognise. The bench’s case ran as follows. Article 21 guarantees that no person shall be deprived of life or personal liberty except according to a procedure that is just, fair and reasonable, and an arrest is among the gravest interferences with liberty the state can make. Where an accusation under the Act could be made unilaterally, registered without any screening, and followed by arrest without any senior officer’s concurrence, an innocent person, and in particular a public servant whose official decisions gave others a grievance against them, stood exposed to a process whose initiation was itself a punishment: arrest, custody, the loss of reputation, and in the case of a government officer the professional consequences that follow an arrest regardless of the outcome. The figures before the Court suggested that this exposure was not theoretical. Cases were being closed as false after investigation, and the proportion of convictions was low. A filter at the point of registration, an approval at the point of arrest, and a judicially administered exception to the bail bar where no prima facie case existed were, on this view, not favours to the accused but the minimum content of a fair procedure, of a kind courts routinely read into criminal statutes. The bench also pointed to the magistrate’s scrutiny of the recorded reasons as a judicial check built into its own scheme, and it made the directions prospective so that concluded matters were untouched.

Three features of the directions, visible on the day they were issued, carried the seeds of the reversal. The first is that none of them interprets any word of the Act. Section 18 excludes Section 438 of the Code; the anticipatory bail conclusion did not construe that exclusion but qualified it by reference to the quality of the complaint. The enquiry and approval directions have no textual anchor in the Act at all; they are machinery, with designated officers, recorded reasons, scrutiny by magistrates and sanctions for breach, which is the characteristic content of legislation and of rules made under legislation. The second feature is that the approval direction placed the power to permit an arrest, in the case of a public servant, in the hands of that servant’s own appointing authority, which in the very class of case that had produced the appeal, an accusation arising out of an official’s own departmental decisions, meant asking the accused’s institution to license the process against him. The third is that the preliminary enquiry, however softly the word “may” was used in the direction, interposed a police assessment of whether allegations were frivolous or motivated between a complainant and the registration of a first information report, in a class of case where the Constitution Bench decision in Lalita Kumari v. Government of Uttar Pradesh, (2014) 2 SCC 1, had held registration to be mandatory on the disclosure of a cognisable offence and had confined preliminary enquiry to defined categories of case, a list in which atrocity complaints do not appear.

The reaction outside the courtroom was immediate and ran in the opposite direction to the judgment’s self-understanding. Organisations of the Scheduled Castes and Scheduled Tribes, and parliamentarians across parties, read the directions not as safeguards for the innocent but as a dilution of the one statute whose strictness was its protection: an enquiry before registration would expose complainants to pressure at the moment of greatest vulnerability, witnesses could be reached and evidence could cool while a Deputy Superintendent assessed motivation, and an approval requirement would hand a veto over arrests to the very administrative hierarchies within which caste power is exercised. In the weeks that followed there were widespread protests across the country, including a nationwide day of protest at the beginning of April 2018 during which lives were lost. This article records that fact without detail and without attributing responsibility for those deaths to any participant or authority; the legal significance of the protests does not depend on their particulars, and their particulars are contested. What is not contested is their political effect. The Union of India filed a petition seeking review of the judgment, and when interim relief was pressed before the same bench in April 2018 the bench declined to stay its directions, observing that they had been misunderstood and that they protected the innocent rather than diluting the Act. The review was in time placed before a bench of three judges. But by then Parliament had already been set in motion, and it moved faster than the Court.

What exactly did the approval direction require?

Two approvals, depending on who was accused. A public servant could be arrested only after approval by the appointing authority; any other person, only after approval by the Senior Superintendent of Police of the district. Each approval required reasons recorded in writing, which the magistrate was to scrutinise before permitting further detention.

That design deserves one further paragraph of analysis, because its weakness was structural rather than incidental. An arrest approval operates before the evidence is tested and at the moment when the investigating officer knows least; it therefore tends to become either a formality, granted on the file as presented, or a veto, exercised for reasons that never reach a courtroom. The Mahajan scheme tried to guard against the veto by requiring recorded reasons and magisterial scrutiny, but scrutiny of reasons for permitting an arrest is not the same as a hearing on whether an offence is disclosed, and the magistrate’s role under the Code at the remand stage is confined to the legality and necessity of custody. The review bench would later fasten on a different and deeper objection, that the direction created two procedural regimes for two classes of complainant, but the practical objection was already available in March 2018 and was made at once by those who work with the Act: a safeguard that operates by delaying registration and arrest protects most reliably the person with the institutional standing to invoke it, which in the run of atrocity cases is the accused and not the complainant.

What the bench did with the authorities against it

A word must be said about the most technically vulnerable step in the March 2018 judgment, because the review and validation judgments both return to it. Balothia had upheld Section 18 as a complete exclusion, and the later cases had applied it as such. The Mahajan bench did not overrule Balothia; a bench of two judges could not overrule a decision of a coordinate bench in any event. It clarified it, holding that the exclusion operated in a case under the Act properly so called, and that where no prima facie case under the Act was made out there was in truth no case under the Act for the exclusion to govern. As a piece of technique this is the threshold reading that Prathvi Raj Chauhan would eventually adopt in a disciplined form, and in isolation it is defensible: an exclusion keyed to an accusation of an offence under the Act has nothing to attach to when the accusation, on its face, is not of such an offence. The difficulty is that the bench did not leave the point in that disciplined form. It joined the threshold reading to an enquiry designed to establish whether allegations were frivolous or motivated, which is an enquiry into truth and motive rather than into the face of the complaint, and to an approval system that operated regardless of the complaint’s facial sufficiency. The judgment thus contained, side by side, a proposition about the construction of Section 18 that the law would in time accept and a machinery of screening that the law would reject, and the public argument of 2018, which treated the judgment as a single act of dilution or of protection, rarely separated the two.

The bench also gave a hostage to fortune in the way it handled the figures. The judgment recorded that the material before it showed a pattern of false implication, and it described the abuse of the law of arrest under the Act as acknowledged. That word, acknowledged, did a great deal of work. It converted a contested empirical claim, contested in the very proceedings by those who intervened to defend the Act’s design, into a premise that the directions could simply apply. When the review came to be argued, the Union’s most effective forensic point was not constitutional but evidential: that the disposal figures of the National Crime Records Bureau, on which the premise rested, record the outcomes of investigations and trials, not findings of falsity, and that a statute’s acquittal rate, in a field where the complainant and the witnesses are structurally exposed to pressure from the accused’s side, cannot be read as a measure of how many complaints were dishonestly made. The review bench did not need to resolve that contest to recall the directions, since its ground was institutional, but the exposure of the premise mattered to the episode’s reception: a judgment that announced itself as a response to an acknowledged abuse was heard, once the acknowledgement was questioned, as a judgment that had taken sides in a factual dispute it had not tried.

Parliament answers: the amendment of August 2018

The Union government’s public position through the spring of 2018 was that the statute would be restored. The Minister for Social Justice and Empowerment, Thaawar Chand Gehlot, whose ministry administers the Act, piloted the government’s response, and Ram Vilas Paswan, then a Union minister and the most prominent Scheduled Caste voice in the ruling alliance, pressed publicly for the judgment’s effect to be undone by legislation if necessary. When Parliament met for its monsoon session, the government chose the legislative route rather than waiting for the review. The Scheduled Castes and the Scheduled Tribes (Prevention of Atrocities) Amendment Bill was passed by the Lok Sabha on 6 August 2018 and by the Rajya Sabha on 9 August 2018. It received the assent of the President on 17 August 2018 and became the Scheduled Castes and the Scheduled Tribes (Prevention of Atrocities) Amendment Act, 2018, Act 27 of 2018. It was brought into force on 20 August 2018.

The Amendment Act is one of the shortest consequential statutes in the modern book. It does a single thing: it inserts a new Section 18A into the Act of 1989. Everything about the episode, the speed, the drafting, and the later judicial treatment, is legible from that section’s text, which is set out and analysed in the next part of this article. For the narrative, what matters is the correspondence between the amendment and the judgment. Each of the three operative directions of March 2018 is answered by a clause of Section 18A, and answered in the negative. The preliminary enquiry is answered by a provision that no preliminary enquiry shall be required for the registration of a first information report against any person. The approval directions are answered by a provision that the investigating officer shall not require approval for the arrest, if necessary, of any person accused under the Act, and that no procedure other than that provided under the Act or the Code shall apply. The anticipatory bail conclusion is answered by a provision that Section 438 of the Code shall not apply to a case under the Act, and this time the exclusion is armoured with words that Section 18 had never carried: “notwithstanding any judgment or order or direction of any Court.”

The legislative technique is worth isolating, because it is the hinge of the doctrinal story. Parliament did not purport to reverse the Mahajan judgment as a judgment; the quashing of the proceedings against Dr Mahajan stood, and no legislature could have disturbed it as between the parties. What Parliament did was to change the underlying law for the future, removing the basis on which the directions rested by enacting the contrary rule in express terms. That is the constitutionally orthodox method of legislative override, and its validity, when the legislature is competent and the new law does not itself offend the Constitution, has never been seriously in doubt in Indian law. A legislature that disagrees with a court’s reading of a statute may amend the statute, and a legislature that disagrees with a court’s directions may enact provisions the directions contradict. What was unusual was not the method but the target and the candour. The notwithstanding clause names no case, but its words “any judgment or order or direction of any Court” were aimed at one judgment, one order and one set of directions, and everyone in the chamber, in the courtroom and in the press gallery knew it. Whether a legislature may go further, and enact that a judgment itself shall be deemed never to have been delivered, is a different and harder question that the amendment carefully did not raise; the boundary of permissible override, and the basic structure limits that stand behind all legislation, are treated in the article on the basic structure doctrine, to which this discussion returns when the constitutional challenge is examined.

Did the amendment only restore the position before March 2018?

In substance yes, with one addition. The enquiry and approval clauses restated the ordinary law the directions had displaced, and the bail clause restated Section 18. The addition was the notwithstanding clause, directed not at accused persons but at courts: the exclusion of anticipatory bail was to operate whatever any court had said about it.

Section 18A: the override in the statute’s own words

Section 18A, headed “No enquiry or approval required”, has two subsections. Subsection (1) provides that, for the purposes of the Act, a preliminary enquiry shall not be required for registration of a first information report against any person, and that the investigating officer shall not require approval for the arrest, if necessary, of any person against whom an accusation of having committed an offence under the Act has been made, and that no procedure other than that provided under the Act or the Code shall apply. Subsection (2) provides that the provisions of Section 438 of the Code shall not apply to a case under the Act, notwithstanding any judgment or order or direction of any Court. The drafting is plain to the point of bluntness, and its three moves repay separate attention, because later litigation tested each of them.

The first move is the phrase “if necessary” in the arrest clause. Section 18A does not direct that every accused person be arrested. It removes the approval requirement and returns arrest to the general law, under which arrest is a discretionary step governed by Sections 41 and 41A of the Code of Criminal Procedure, and by the discipline the Supreme Court laid down in Arnesh Kumar v. State of Bihar, (2014) 8 SCC 273, requiring an officer to satisfy himself on a statutory checklist that arrest is needed for investigation, for preventing further offences, for preventing the disappearance of evidence or influence over witnesses, or for securing the accused’s presence, and to use a notice of appearance where it is not. That arrest discipline, which operates on the officer’s satisfaction and is examinable by the magistrate at remand, was never touched by Section 18A, a point of considerable practical importance that much commentary on the amendment missed. The comparison between the old Code and its successor codes, including where the arrest provisions and the anticipatory bail provision sit in the successor code, is maintained in the comparison of the Indian Penal Code and the Bharatiya Nyaya Sanhita; the anticipatory bail provision of Section 438 of the Code of Criminal Procedure is carried into Section 482 of the Bharatiya Nagarik Suraksha Sanhita, 2023, and the substance of the exclusion that Section 18 and Section 18A enact is unaffected by the change of code.

The second move is the closing clause of subsection (1): no procedure other than that provided under the Act or the Code shall apply. This is the anti-direction clause in its general form. It forbids not merely the Mahajan machinery but any procedure interposed from outside the two named sources of law, and it does so prospectively against courts, commissions and police standing orders alike. Read with subsection (2), it converts the interpretive question that had consumed the spring of 2018 (what may a court add to this statute?) into a question the statute itself answers: nothing that is not in the Act or the Code. Whether Parliament can bind the constitutional courts in this way when they exercise their constitutional jurisdiction, as distinct from their statutory appellate jurisdiction, is a subtler matter, and the 2020 judgment is careful with it, as will appear. But as a statement of the ordinary law governing police stations, magistrates and sessions courts, the clause is complete.

The third move is the notwithstanding clause itself, and it is the phrase this article was commissioned to quote: the provisions of Section 438 of the Code shall not apply to a case under the Act “notwithstanding any judgment or order or direction of any Court”. Non obstante clauses of this breadth are familiar in fiscal and regulatory statutes, where they override other enactments. Directed at judgments, the clause is rarer and more pointed. Its legal effect is not to annul any judgment; no statute can annul the decision of a case between parties. Its effect is to remove the precedential and operational basis of the directions for the future: after 20 August 2018, a sessions judge asked for anticipatory bail in a case under the Act was confronted not merely with Section 18 as construed in Balothia but with a fresh legislative command that the exclusion operated whatever any court had directed. Practically, the clause did its work in the five months between the amendment and the recall, because it gave every subordinate court a statutory ground to refuse anticipatory bail and to disregard the enquiry and approval requirements without waiting for the Supreme Court to revisit Mahajan. Constitutionally, it raised the question the petitioners in Prathvi Raj Chauhan would put in the sharpest form: whether a legislature may, by a notwithstanding clause aimed at a judgment, trench upon the judicial function itself. The Court’s answer in February 2020 was that it may, where what it enacts is a change in the law within its competence, and that the clause changed the law rather than deciding a case.

Why the amendment is a single section

The form of the amendment deserves a paragraph of its own, because it was a choice among visible alternatives and the choice shaped everything that followed. The government might have reopened the Act generally, revisiting the catalogue of offences, the special courts and the relief machinery, as the Amendment Act of 2015 had done. It did the opposite. The Statement of Objects and Reasons accompanying the Bill identified the judgment as the occasion of the legislation and the restoration of the pre-judgment position as its object, and the Bill confined itself to the one section required for that object. Three consequences followed from the minimalism. First, the parliamentary debate could be, and was, confined to the single question whether the directions should stand, which suited a government legislating in a charged atmosphere and against a session calendar. Second, the constitutional challenge, when it came, had a correspondingly narrow target: the petitioners could attack Section 18A, but they could not easily convert the case into a general audit of the Act, whose core had been upheld a generation earlier. Third, and most subtly, the minimalism strengthened the government’s hand in the review, because it demonstrated that the legislative branch had responded to the Court in the most restrained form available to it, changing no more law than the judgment had displaced. A legislature that answers a direction with a single section, rather than with a general revision of the field, presents itself to a reviewing court as a coordinate branch correcting a specific error, and the courts in both the recall and the validation treated the amendment in exactly that spirit. Legislative minimalism is not always a virtue; here it was the amendment’s chief strategic asset.

Twenty-three months, four positions

The sequence is usually told in slogans, as dilution followed by restoration, or as protection followed by surrender. Told step by step, with what a person accused under the Act could and could not do at each stage, it is more instructive and less dramatic: the law changed four times in twenty-three months, and at no stage was the position the one either slogan describes.

Step and date	Institution acting	Legal position immediately afterwards	What an accused person could and could not do
Directions, 20 March 2018	Supreme Court, two-judge bench in Mahajan	Anticipatory bail not barred outright where no prima facie case or mala fide complaint; arrest of a public servant only with the appointing authority’s approval, of others only with the Senior Superintendent’s approval; a Deputy Superintendent could hold a preliminary enquiry before registration	Could apply for anticipatory bail on the threshold ground; could not be arrested without the recorded approval applicable to their class; could face registration only after any enquiry the Deputy Superintendent chose to hold
Protests, early April 2018	Organisations of the Scheduled Castes and Scheduled Tribes, with support across parties	The directions remained the law; the Union of India sought review and pressed for, and was refused, a stay; Parliament was committed to legislate	Position unchanged in law; the practical certainty of the directions was already weakened by the announced review and the promised amendment
Amendment, 20 August 2018	Parliament, by Act 27 of 2018 inserting Section 18A	No preliminary enquiry required before registration; no approval required for arrest; Section 438 of the Code inapplicable notwithstanding any judgment, order or direction of any court	Could not obtain anticipatory bail under the amended text; could not insist on an enquiry or an approval; retained arrest safeguards under the Code itself and the remedies of bail after arrest and quashing
Recall, 1 October 2019	Supreme Court, three-judge bench on review in Union of India v. State of Maharashtra	The approval and enquiry directions of March 2018 recalled as contrary to the statute and beyond the judicial function; the directions ceased to operate as a source of rights	Governed wholly by the statute as amended; the directions ceased to be a source of rights in any pending or future matter
Validation, 10 February 2020	Supreme Court, three-judge bench in Prathvi Raj Chauhan	Section 18A upheld as constitutional; where a complaint on its face does not disclose an offence under the Act, the bar on anticipatory bail does not bite and the ordinary jurisdiction of the courts remains	Could apply for anticipatory bail only on the ground that the complaint discloses no offence under the Act at all; could seek quashing of a complaint that discloses no offence; could not obtain anticipatory bail as a matter of course

The table discloses the feature of the sequence that slogans conceal. The two extreme positions, the full Mahajan filter and the unqualified statutory bar, each governed for only a matter of months. The position that has endured is the one produced by the last step, in which the statute as amended by Parliament was upheld by the Court with a clarification that the Court, not Parliament, supplied, and which neither the amendment’s supporters nor its critics had asked for in that form. That is the position analysed after the recall and validation judgments are examined in turn.

Twenty months of overlapping regimes

Between March 2018 and February 2020 the governing text changed faster than the institutions applying it could comfortably follow, and a short account of the overlap explains much of the confusion in the commentary of the period. From 20 March to 20 August 2018 the Mahajan directions were the law declared by the Supreme Court, and police stations and courts were bound by them: enquiries were held, approvals were sought and recorded, and anticipatory bail was granted and refused on the threshold test the judgment had announced. Yet throughout those five months the position was visibly provisional. The Union’s review petition, Review Petition (Criminal) No. 228 of 2018, was pending, the deciding bench had declined interim relief while making its displeasure at the public characterisation of its judgment plain, and the government had announced its legislative intention. A direction expressed to operate prospectively, issued by a two-judge bench against the text of a statute, and immediately disowned by the executive and the legislature, occupied about as insecure a position as a proposition of Indian law can occupy while remaining formally binding.

The amendment then superimposed the statutory regime without immediately removing the judicial one. From 20 August 2018, a subordinate court asked to apply the enquiry or approval directions was confronted with Section 18A’s command that no such procedure applied, armoured by the notwithstanding clause; the directions remained unrecalled on the Supreme Court’s own record, but their statutory foundation had been removed, and in practice the amendment governed. The challenges to Section 18A, of which Prathvi Raj Chauhan’s writ petition, Writ Petition (Civil) No. 1015 of 2018, became the lead case, were filed within weeks, and the Court issued notice on them in September 2018. The Union’s stance in that litigation completed the alignment against the directions: the same government that had sought their review defended the amendment enacted to displace them, filing an affidavit that defended Parliament’s competence and resisted the use of acquittal rates to infer that complaints under the Act were generally false, a submission whose forensic significance has already been noted. In January 2019 the challenges and the review petitions were clubbed together and the review was referred to a bench of three judges, so that the fate of the directions and the fate of the amendment would be decided by benches of the same composition in sequence. That sequencing, recall first and validation four months later, is what gives the completed episode its unusual coherence: the validation bench could treat the directions as already withdrawn and address itself to the statute alone, and the recall bench could reason about judicial power without needing to lean on the amendment that had overtaken its work.

Steps taken under the directions: the transition problem

A sequence in which the law changes four times in twenty-three months generates a class of case that none of the four steps was designed to address: the matter that was partway through the process when the ground moved. Complaints were enquired into, and some were not registered, during the five months when the directions operated. Arrests were approved and made, and anticipatory bail was granted, under a regime that was later displaced and then disowned. Accused persons were refused anticipatory bail by courts applying Section 18A while the directions formally remained unrecalled. The handling of that residue followed from the character of each step rather than from any transitional code. The directions had been made prospective, so steps taken under them were taken under the law as it then stood and were not invalidated by their recall; the recall withdrew a rule for the future, it did not falsify the past. The amendment was likewise prospective in its operation on procedure, and it neither annulled bail granted nor revived complaints that an enquiry had discouraged, some of which will simply never have been recorded anywhere, which is among the reasons the episode’s true incidence cannot be reconstructed from the registers. And the validation, in preserving the threshold jurisdiction, gave courts a tool for the live cases without purporting to unsettle concluded ones. The transition was therefore untidy in the way legal transitions usually are, clean at the level of doctrine and uneven at the level of the individual file, and the unevenness fell, as it generally does, on the parties least able to litigate their way out of it. It is a further reason for the institutional moral the recall drew: a court that legislates prospectively for a class of live cases assumes a transitional responsibility that courts are structurally ill equipped to discharge, since a judgment cannot contain the savings and commencement provisions by which a legislature manages the passage from one regime to another.

The recall: the Court disowns its directions

The review petitions against the Mahajan judgment were decided on 1 October 2019 in Union of India v. State of Maharashtra, reported at (2020) 4 SCC 761, by a bench of Arun Mishra, M. R. Shah and B. R. Gavai JJ. The Union of India, which had sought the review, argued through the Attorney General that the directions had rewritten a protective statute, that they rested on an impermissible use of the Court’s power under Article 142 of the Constitution to pass orders for doing complete justice, and that they discriminated in substance against the very class the Act protects by making offences against that class harder to register and slower to act upon than offences against anyone else. The bench accepted the substance of that case. It held that the approval and enquiry directions issued in Mahajan deserved to be and were recalled, and that the direction attaching disciplinary and contempt consequences to their breach fell with them.

The reasoning repays attention because it is the doctrinal centre of the whole episode. The bench held, first, that the directions encroached upon the field reserved for the legislature. The Act was not silent in a way that invited supplementation; its scheme was a considered one, and to require an approval before arrest and an enquiry before registration was to add conditions for the operation of the Act which the legislature had not provided and which altered the Act’s working in its most sensitive area. Second, the bench held that the directions were impermissible within the parameters the Court has laid down for the exercise of its Article 142 power, drawing on the line of authority, including Supreme Court Bar Association v. Union of India, (1998) 4 SCC 409, which holds that the power to do complete justice operates within the law and cannot be used to supplant substantive statutory provisions or to build a new edifice where the legislature has built one. Third, the bench reasoned in terms of the Constitution’s own commitments: the directions sat uneasily with the concept of protective discrimination in favour of the downtrodden classes under Article 15(4), because their practical burden fell on complainants under a statute enacted for the protection of those classes, and a procedure that is formally neutral as between complainants is not neutral in effect when it applies only to complaints by members of the protected class. Fourth, the bench treated the preliminary enquiry direction as inconsistent with Lalita Kumari, where a Constitution Bench had confined the permissible categories of pre-registration enquiry and had held registration mandatory on disclosure of a cognisable offence; a two-judge bench could not, consistently with that authority, open a new category for the class of complaints where delay is most damaging.

Two things the recall did not do are as significant as what it did. It did not hold that the Mahajan bench had been wrong to quash the proceedings before it; the abuse of process conclusion was not the target, and the review was directed to the general directions. And it did not rest its conclusion on the 2018 amendment. The recall was reasoned from the Constitution and from the limits of judicial power, not from the fact that Parliament had already removed the directions’ statutory basis. That choice of ground is what elevates the case from a footnote to the amendment into the leading modern statement of the interpretation and legislation boundary: the Court held that its own directions would have been impermissible even if Parliament had never acted. The amendment and the recall are thus independent corrections arriving at the same place from opposite directions, the legislature removing the directions’ foundation and the judiciary withdrawing the directions themselves, and the convergence is the reason the episode is taught as a case study rather than as a controversy.

The register of the review judgment also deserves notice, because it differs from the institutional prose in which such corrections are usually delivered, and the difference was part of the message. The bench wrote at length about the historical reality the Act addresses: the persistence of caste-based humiliation and violence, the constitutional framers’ expectation that such legislation would one day be unnecessary, and the distance between that expectation and the record. It closed with the expressed hope that a day would come when no such legislation, and no reservation, would be required because only one class of human beings existed, equal in all respects. Passages of that kind in a judgment recalling directions serve a function beyond rhetoric. They tell the audience most affected by the correction, the communities whose organisations had protested the directions, that the Court’s institutional self-correction is not a retreat from the statute’s purpose but a defence of it, and they tell the audience on the other side that the grievance about abusive complaints, which the judgment does not deny, must find its remedy in a forum other than this one. A reader comparing the three principal documents of the sequence will notice that each speaks most directly to a different constituency: the March 2018 judgment to the accused and the official, the amendment to the complainant and the community, and the review judgment to both at once, which is some part of why it succeeded in closing a controversy that the first two documents, each in its own way, had opened.

On the evidential question the review bench was careful in a way that repays emulation. It did not hold that complaints under the Act are never false, and it did not disturb the quashing in Mahajan itself, which rested on the particular facts. What it resisted was the inference from outcomes to falsity. A case that ends in acquittal may have been false, or weak, or true and unprovable in the conditions in which it had to be proved, and the bureau’s tables do not distinguish among those possibilities. The bench treated the figures accordingly, as incapable of bearing the weight the directions had placed on them, and it reinforced the point from the structure of the Act: a legislature that creates special courts and victim protections because it expects the ordinary process to fail a class of complainant has already made its assessment of why that class’s cases end as they do, and a court displacing that assessment needs better material than the assessment’s own subject matter, unexamined. That is not a holding that could be codified, but as a standard for the use of statistics in constitutional adjudication it is the most transferable proposition in the entire sequence.

Is arrest automatic once an FIR is registered under the Act?

No. Section 18A removes the approval requirement and the enquiry, and nothing in it directs an arrest. Arrest remains governed by the Code, under which an officer must be satisfied that arrest is necessary for the purposes the Code specifies, record reasons, and use a notice of appearance where arrest is not necessary.

That answer matters because it locates the surviving safeguard correctly. The protection against unnecessary arrest in a case under the Act is the general arrest discipline of the criminal process, enforced by magistrates and by the High Courts, and not any filter peculiar to the Act. The March 2018 bench thought that discipline insufficient for this class of case and built a second one on top of it. The review bench’s position, carried into the validation, is that a second filter peculiar to this Act is not the Court’s to build and was not Parliament’s choice, and that the remedy for an arrest that the general discipline should have prevented is the remedy the general law supplies: release on bail, and proceedings against a demonstrably false accusation under the provisions that penalise it.

The validation: Prathvi Raj Chauhan and the precise position on bail

The constitutional challenge to Section 18A was decided on 10 February 2020 in Prathvi Raj Chauhan v. Union of India, reported at (2020) 4 SCC 727, by a bench of Arun Mishra, Vineet Saran and S. Ravindra Bhat JJ. Arun Mishra J. wrote the principal opinion for himself and Vineet Saran J.; S. Ravindra Bhat J. wrote a separate concurring opinion. The petitioners’ case, in its strongest form, was that Section 18A violated Articles 14 and 21: it excluded anticipatory bail altogether for a class of accusation defined not by the gravity of the offence but by the identity of the complainant, it removed every procedural filter at the registration and arrest stages, and its notwithstanding clause was aimed at the judiciary in a manner that offended the separation of powers. The Union defended the amendment as a competent exercise of legislative power restoring a protective scheme whose validity, in its Section 18 core, had already been upheld in Balothia, and as a permissible legislative response to a judicial direction, of the kind legislatures routinely make.

The Court upheld Section 18A. Three strands of the principal opinion explain the result. The first is competence and override. The amendment did not decide any case and did not purport to annul the Mahajan judgment; it changed the law prospectively within Parliament’s legislative field, and a legislature may do that in response to a judgment it considers wrong, including a judgment of the Supreme Court, provided the new law is itself constitutional. The notwithstanding clause was, on this reasoning, an emphatic way of changing the law and no more. The second strand is the protective scheme. Reading the Act against Articles 15, 17 and 21, and against the constitutional commitment to abolish untouchability and to protect the Scheduled Castes and Scheduled Tribes from social injustice and exploitation, the Court treated the stringency of the procedure as a legislative judgment about the conditions in which atrocity complaints are made, including the exposure of complainants and witnesses to pressure, which the Court was not positioned to second-guess on the material before it. The third strand is the one that gives the judgment its practical importance, and it is a strand of limitation rather than of validation. The Court observed that the provisions of Section 18A had in large measure been rendered academic by the recall of the Mahajan directions, since they were enacted to take care of a mandate that no longer prevailed; and it then stated the qualification on which the practical law rests.

The qualification must be stated exactly, because the brief for this article flags it as the proposition most often overstated in both directions, and the flag is justified. The Court held that the bar on anticipatory bail in Sections 18 and 18A operates where a case under the Act is made out. Where the complaint, on its face, does not disclose an offence under the Act at all, the bar does not come into play, because there is no case under the Act to which the bar can attach, and in that situation the ordinary jurisdiction of the courts is not excluded: a court may grant anticipatory bail, and the power to quash a proceeding that discloses no offence remains intact. Where the complaint does disclose the ingredients of an offence under the Act, anticipatory bail is not available, and it cannot be obtained by inviting the court at the bail stage to weigh the evidence, to assess the credibility of the complainant, or to conduct the mini-trial that the threshold enquiry is not. The enquiry is into the face of the complaint, not into its truth. S. Ravindra Bhat J., concurring, put the same position in more guarded terms: anticipatory bail in a case under the Act is exceptional and is not to be granted in the ordinary course, and the occasion for it is the case in which no prima facie offence is disclosed; his opinion added a reflection, absent from the principal judgment, on the constitutional value of fraternity and on the equal citizenship that the Act exists to make real, which he treated as a consideration of the same order as the liberty of the accused rather than as a makeweight against it.

The concurrence deserves separate attention because it supplies the register in which the validated position is most likely to be applied by the courts below. S. Ravindra Bhat J. approached the case from the constitutional promise of equal citizenship, and from the Preamble’s fraternity, which he treated not as decoration but as an operative value in a case where the liberty of one citizen was being weighed against the dignity and security of another. On the operative question he was, if anything, more restrictive than the principal opinion: anticipatory bail, in his formulation, is to be granted sparingly and in exceptional circumstances, the paradigm being the complaint that does not make out a prima facie case under the Act, and he warned against the conversion of the threshold enquiry into a general evaluation of the prosecution’s prospects. The practical consequence of reading the two opinions together, as a court below must, is that the threshold jurisdiction survives with a deliberately narrow gate: the applicant must show, from the complaint itself, the absence of an ingredient of the offence, and not merely the weakness of the evidence, the improbability of the allegation, or the hardship of arrest. Counsel who treat Prathvi Raj Chauhan as a general softening of the bar misread the temper of both opinions, and the steady refusal of the High Courts, in the run of cases decided on the validated position, to convert the threshold into a merits preview is the concurrence’s logic working itself out in practice.

It follows that the two statements in widest circulation are both wrong, and wrong in a way that matters to anyone advising on a live case. The statement that anticipatory bail is available in atrocities cases, full stop, ignores that the bar operates wherever the complaint discloses an offence, which is most cases that reach a court. The statement that anticipatory bail is barred in every case, full stop, ignores that the bar attaches to a case under the Act, and that a complaint which does not disclose the Act’s ingredients, for example because the alleged insult was not in public view, or because the accusation does not allege the caste-based character the relevant clause of Section 3 requires, or because the dispute on its face is of a kind the Act does not reach, is not such a case. The space between the two statements is narrow, and deliberately so, but real cases are decided in it, and the work of counsel at the threshold is the work of showing, from the four corners of the complaint, either that the ingredients are present or that they are absent.

Did the February 2020 judgment bring back the Mahajan safeguards?

No. The enquiry and the arrest approvals were not restored in any form; the judgment upheld the amendment that abolished them. What it preserved was not the Mahajan machinery but the ordinary threshold jurisdiction of a court to ask whether the complaint before it discloses an offence under the Act at all.

What remedies remain for a person facing a complaint that discloses no offence under the Act?

Three, at different points. Before arrest, anticipatory bail on the ground that the complaint on its face discloses no offence under the Act. At any stage, a petition to the High Court to quash a proceeding that discloses no offence. After arrest, regular bail on ordinary principles, which Sections 18 and 18A do not touch.

The quashing route, under the High Court’s inherent power to prevent abuse of the process of any court or under Article 226 of the Constitution, is the very route by which the Mahajan case itself reached the Supreme Court and was decided, and its survival through every stage of the sequence is the strongest single evidence that the validated regime is not the regime of unfiltered process its critics describe. A complaint that discloses no offence under the Act was always liable to be quashed, before 2018, during the directions, and after the validation; what changed across the sequence was never the existence of that jurisdiction but the stage at which, and the machinery by which, complaints were to be screened before it could be invoked.

If Section 18A became academic, why does it matter?

The most quoted sentence of the validation judgment is also the most misunderstood. When Arun Mishra J. observed that the provisions of Section 18A had been rendered of academic use because they were enacted to take care of a mandate that no longer prevailed, he was explaining why the challenge to them had lost its practical edge after the recall, not holding that the section had become a dead letter. The distinction matters, because Section 18A continues to do at least four kinds of work that the recall alone could not have done. It binds every subordinate court, police officer and magistrate as statutory text, whereas the recall binds as precedent; a direction-seeking litigant who might have argued around a precedent’s edges confronts, in the section, a text that admits no edges. It forecloses the future in a way a judgment cannot: the recall states the law as a bench of three judges declared it, and a later bench of the Supreme Court could in principle revisit it, whereas no court can revisit a competent statute except on constitutional grounds, which the validation has closed off for this provision. It speaks, in its closing words, to procedures “other than that provided under this Act or the Code” from any source, which catches administrative improvisation by police headquarters and state circulars as well as judicial directions, a field a judgment about judicial power does not reach. And it records, in the permanent form of the statute book, the considered position of the democratic branch on a question of contested social policy, which is the form in which such settlements are supposed to be recorded in a system where courts decide cases and legislatures decide policy.

The observation about academic use also contains a quiet warning in the opposite direction, which commentary has mostly missed. If a provision enacted to displace a judgment becomes academic when the judgment is withdrawn, then a provision enacted to displace a judgment that is not withdrawn retains its full operative force, and its validity will be tested entirely on constitutional grounds. The 2018 amendment survived that test for the reasons the validation gave. But the episode establishes the template for the harder case: a legislature that overrides a judgment the Court stands by, on a subject where the constitutional arguments are closer, will find in Prathvi Raj Chauhan both its best authority, that changing the basis of a judgment is legislation, and its clearest limit, that the changed basis must itself survive the Constitution. The academic remark is the hinge between the two halves of that template, and it is likely to be quoted in override litigation long after the facts of this sequence have faded from memory.

The threshold question in concrete cases

The practical doctrine that emerges from Prathvi Raj Chauhan is best fixed by running it against the recurring patterns, because the threshold enquiry is a discipline of reading complaints, and its content becomes clear only when it is seen deciding things. Take first the pattern that generated the whole sequence: a complaint against a public servant for something done in an official capacity, of which the sanction refusal in Mahajan is the archetype, and complaints against officers for adverse reports, for demolition or eviction drives, for examination and selection decisions, and for disciplinary orders are the common variants. Under the regime as validated, such a complaint is registered without enquiry and the officer has no approval to shelter behind. But the threshold question remains fully alive: an official act, without more, is not an offence under the Act, and unless the complaint alleges facts that bring it within a clause of Section 3, for instance an intentional insult or intimidation with intent to humiliate, committed in a place within public view, the complaint discloses no offence under the Act and both anticipatory bail and quashing are available. This is, in substance, the result Mahajan reached on its own facts, reached by the route the validated law approves.

Take next the pattern of the caste-name abuse alleged to have occurred inside a house or another place not open to public view. The clauses of Section 3, as recast by the Amendment Act of 2015, that penalise intentional insult and abuse by caste name both carry the requirement that the act occur in any place within public view, a requirement Parliament wrote into the text and the courts have enforced as written. In Hitesh Verma v. State of Uttarakhand, decided in November 2020 and reported at (2020) 10 SCC 710, the Supreme Court quashed proceedings where the alleged abuse occurred inside the complainant’s house, holding that the public view ingredient was absent and that a property dispute had been given the colour of an atrocity complaint. The case is the threshold doctrine at work after the validation, and it illustrates the doctrine’s precise character: the Court did not assess whether the abuse happened, or whether the complainant was truthful, or whether the parties’ larger dispute was genuine. It asked whether the complaint’s own allegations, if true, amounted to the offence, and finding an ingredient missing on the face of the allegations, it ended the proceedings. That is a judicial function as orthodox as any in the criminal law, and the fact that it disposes of cases the complainant experiences as meritorious is the standing grievance of the other side: the ingredients of the clauses, and in particular the public view requirement, exclude conduct that is humiliating in substance, and the boundary of the Act’s coverage is set by those ingredients rather than by the threshold doctrine, which merely enforces them.

Take finally the pattern that the validation leaves most exposed: the complaint that does disclose every ingredient on its face and is nevertheless, as the investigation later shows, untrue in fact. For that accused there is no anticipatory bail, no enquiry and no approval, and the remedies are arrest, remand scrutiny, regular bail and trial, with the possibility of proceedings for a false complaint only after the falsity is established in the manner the general law requires. The March 2018 bench regarded that exposure as the central injustice of the scheme. Parliament, in enacting Section 18A, and the Court, in upholding it, jointly took the view that the exposure is the price of a registration regime that does not pre-judge complainants, and that the class of cases in which the price is paid by an innocent person is the class the general law’s post-arrest remedies exist to relieve. A reader is not obliged to find that answer satisfying. It is, however, the answer the constitutional process produced, and any account of the present law that softens it in either direction misstates the position that the four steps, taken together, established.

The mirror image: the cruelty provision sequence

The series has met this institutional pattern once before, in reverse, and the brief for this article directs that the two sequences be read together. The comparison is genuinely illuminating, because the same judicial instinct produced a screening mechanism in both fields, and the fate of the mechanism differed in a way that isolates the variable that decides these cases. The full narrative belongs to the article on the judicial response to Section 498A, and is compressed here to the points of comparison.

In the cruelty provision governing a husband and his relatives, the sequence ran as follows. In Arnesh Kumar in 2014, the Supreme Court disciplined arrest by enforcing the arrest conditions that the Code of Criminal Procedure itself contains, a step nobody of consequence has since disputed. In Rajesh Sharma v. State of Uttar Pradesh, decided on 27 July 2017 by the same two judges who would decide Mahajan eight months later, the Court went further and directed the constitution of Family Welfare Committees in every district, to which every complaint was to be referred, with no arrest ordinarily to be made until the committee reported. In Social Action Forum for Manav Adhikar v. Union of India, decided on 14 September 2018, a three-judge bench set the committee directions aside, holding that the creation of such a mechanism was not within the judicial function and amounted to legislation in an area the statute occupied, while expressly preserving the Arnesh Kumar discipline on the ground that it enforced the Code rather than adding to it. The mechanism then returned by a route nobody predicted: a High Court framed guidelines of its own, and years later the Supreme Court approved those guidelines and directed that they be followed, acting under its power to do complete justice in the case before it and without overruling the 2018 holding.

Set the two sequences side by side and three contrasts emerge. The first is the direction of the first correction. In the cruelty field the correction of the judicial mechanism came from the judiciary alone, in 2018, and Parliament never legislated on the committees at all. In the atrocities field the correction came from the legislature first, in August 2018, and from the judiciary second, in October 2019, and the order matters: because Parliament had already enacted the contrary rule, the review bench in the atrocities case could recall the directions against the background of a democratic decision to the same effect, whereas the 2018 bench in the cruelty case had to carry the whole weight of the institutional argument itself. The second contrast is the ground of objection. In both fields the objection that prevailed was identical in form: the mechanism was not in the statute, and creating it was legislation. But in the cruelty field the mechanism was a new body of lay persons interposed in the process, while in the atrocities field the mechanism was a set of official approvals and an enquiry inside the existing police hierarchy; the atrocities directions were, if anything, the more modest invention, and they fell faster and more completely, which suggests that the decisive factor was not the size of the invention but the presence of a legislature willing to answer it. The third contrast is the ending. The cruelty sequence ends in doctrinal untidiness, with a mechanism set aside as beyond the judicial function in 2018 operating under judicial authority years later through a different door, and the 2018 holding left standing beside it. The atrocities sequence ends in doctrinal tidiness of an almost diagrammatic kind: direction, override, recall, validation, each step taken by the institution constitutionally placed to take it. Which ending is better is a question this article takes up in the assessment that follows, because it is not obvious that tidiness is the same thing as justice in the individual case, and the accused person whose complaint discloses no offence cares about the threshold remedy, not about the architecture.

How long did each stage of the atrocities sequence govern the law?

The Mahajan directions governed from 20 March 2018 until the amendment displaced them on 20 August 2018, five months. Section 18A governed alone from August 2018 until the recall of 1 October 2019 confirmed its basis from the judicial side. The validated and clarified position dates from 10 February 2020, twenty-three months after the directions were issued.

Filters Parliament itself has written elsewhere

It would be a misreading of the whole episode to conclude that Indian criminal law disapproves of procedural filters, or that the objection to the Mahajan directions was an objection to screening as such. Parliament writes filters into criminal statutes constantly, and several of them are more severe than anything Mahajan proposed. The anti-terror statute conditions bail itself on the court being satisfied, from the case diary and the police report, that there are no reasonable grounds for believing the accusation to be prima facie true, a filter that operates after arrest and has been sustained. The narcotics statute subjects bail to twin conditions concerning the prosecution’s opportunity to oppose and the court’s satisfaction about innocence and future conduct. The general law requires the sanction of a superior authority before a public servant is prosecuted for acts done in the discharge of official duty, a device whose purpose is precisely the one Mahajan invoked, the protection of officers from vexatious prosecution for official decisions, and the corruption statute contains its own sanction requirement for the same reason. The Code of Criminal Procedure itself, in the categories identified in Lalita Kumari, permits a preliminary enquiry before registration in classes of case, matrimonial and family disputes, commercial offences, medical negligence, corruption cases and cases of abnormal delay among them, where the Constitution Bench thought the bare disclosure test needed supplementing.

The comparison yields the episode’s most precise lesson. Every one of those filters was enacted by a legislature in the text of a statute, after whatever deliberation the legislative process supplied, and each has been applied by courts as written, including where its operation is harsh. The filters that failed in 2018 were not weaker than these; the arrest approval Mahajan required is a close cousin of the sanction requirement Parliament itself maintains elsewhere. They failed because of their authorship and their target: authored by a court, for a statute whose legislature had declined, two years earlier, to write them, and aimed at the class of complaint the statute exists to encourage. The lesson for any future bench tempted to repeat the experiment is therefore not that screening is illegitimate. It is that screening is legislation, that its legitimacy depends on the enacting institution, and that a court which believes a statute needs a filter has exactly one constitutionally clean way to say so, which is to say so, in a judgment that decides the case before it and commends the reform to Parliament. The Mahajan bench had that course open to it, and the case before it, which it decided correctly on the abuse of process ground, would have carried the recommendation with considerable force. That it chose the other course is what converted a judgment that might have changed the law through persuasion into a judgment that changed it back through reaction.

The law of legislative override, and why this override was the easy case

The 2018 amendment belongs to a recognised family of Indian legislation, and placing it in the family clarifies why its validity was never in real doubt once the recall had been decided. The family’s governing distinction, built up over decades of validation Act litigation, is between a legislature that changes the law on which a judgment rests and a legislature that merely declares the judgment wrong. The first is legislation. The second is an exercise of judicial power by a body that does not hold it, and it fails. The classic statements come from the validation cases, in which levies collected under a taxing provision that a court had invalidated were sought to be saved: the legislature could re-enact the levy with the defect removed and validate the collections retrospectively, because the judgment’s basis, the defective provision, had been replaced by a competent one, as the Court explained in Indian Aluminium Company v. State of Kerala, (1996) 7 SCC 637, drawing on the earlier validation jurisprudence. The limiting case on the other side is the Cauvery reference, In re Cauvery Water Disputes Tribunal, 1993 Supp (1) SCC 96(2), where a State ordinance that purported to nullify an interim order of a tribunal constituted under a central Act was held unconstitutional: the ordinance did not change any underlying law but simply declared the adjudication ineffective, which is the one thing a legislature may not do to a judgment or an award.

Measured against that distinction, the 2018 amendment is comfortably on the permissible side, and Prathvi Raj Chauhan’s treatment of it becomes predictable. Parliament did not declare the Mahajan judgment ineffective. It changed the underlying law, prospectively and in general terms, by inserting a section that prescribes the procedure for every case under the Act, and the fact that the change was motivated by disagreement with a judgment, and drafted so that its target could not be mistaken, does not convert legislation into adjudication. Legislatures respond to judgments; that is one of the ordinary feedback loops of a constitutional system with judicial review, and the loop runs in both directions and in every field, from tax to land reform to criminal procedure. The petitioners’ more interesting argument was that the notwithstanding clause crossed the line by naming the judiciary as the thing overridden. The Court’s implicit answer, consistent with the validation cases, is that a non obstante clause identifies the material the new law displaces, and judgments and directions that state the displaced law are properly within its reach; the clause overrides the proposition of law, not the decision of the case. Dr Mahajan’s quashed proceedings stayed quashed, and no accused person whose anticipatory bail had been granted under the directions was made liable to rearrest by the amendment, which is the practical test of the distinction and which the amendment satisfied.

The ease of the override question in this episode should not be allowed to obscure the harder variant that the sequence avoided. Suppose Parliament, instead of changing the procedure, had enacted that the Mahajan directions “shall be deemed never to have been issued” while leaving the statutory text untouched, or had re-enacted Section 18 verbatim with a declaration that the Court’s construction of it was erroneous. The first formulation decides nothing about the law and merely annuls an act of the Court, and on the Cauvery reasoning it would have been vulnerable. The second is closer to the line: a legislature may always re-enact a provision, but a declaration about the meaning of an unchanged provision is an instruction to courts about interpretation, and the Indian authorities treat such declarations as effective only where the legislature couples them with an alteration in the law’s basis. The 2018 draftsman avoided both traps by the simple expedient of writing new law. There is a lesson in that for the study of legislative drafting: the override that works is the one that behaves like legislation, general, prospective in substance, and addressed to the rule rather than to the ruling.

What the sequence settles about interpretation and legislation

The strongest case for the March 2018 directions was never really answered on its own terms, and candour requires it to be put once more at full strength. It is this. Arrest is the point at which the criminal process inflicts its most irreversible harm, and it inflicts that harm before any court has examined the accusation. A legal system that permits arrest on an unexamined accusation, in a class of case where accusation is cheap and the social and professional consequences of arrest are severe, has a defect at its centre, and the defect does not disappear because the statute that permits it is a protective one. Courts read procedural fairness into criminal statutes as a matter of routine; the entire modern law of arrest, of remand, of bail and of speedy trial is judge-made in this sense, built on the foundation of Article 21 over decades, and much of it consists of conditions that no legislature enacted in terms. On this view the Mahajan directions were continuous with that tradition, modest in comparison with the arrest jurisprudence the Court had already built, and directed at a documented pattern rather than at a hypothetical one. The bench that issued them included a judge who had spent years on the problem of arrest abuse, and the directions carried the magistrate’s scrutiny as an internal judicial check. Their defenders are entitled to add that the recall judgment did not deny the pattern of abusive accusation; it denied only the Court’s authority to answer it in that way, which is a concession that the grievance was real and remains unremedied by the institution that identified it.

The strongest case against the directions, and for the amendment and the recall, is equally formidable, and on the institutional question it is the case that prevailed. It runs as follows. Interpretation is the ascertainment of the meaning of a text the legislature enacted. The Mahajan directions did not ascertain the meaning of any text; no party contended that Section 18 was ambiguous, and the enquiry and approval machinery has no foothold in any section of the Act. What the bench did was to enact, for one statute, a short procedural code of its own, complete with designated officers, recorded reasons, supervisory scrutiny and sanctions for breach, and to do so in a field where the legislature had legislated comprehensively two years earlier and had chosen the opposite design. The vice was compounded by the design chosen. An approval requirement for the arrest of a public servant lodges the decision in the accused’s own appointing authority, which is to say in the institution whose conduct is frequently the substance of the complaint; a pre-registration enquiry into whether a complaint is motivated asks the police to adjudicate the complainant’s credibility before the machinery of investigation, with its protections for witnesses, has begun to run; and both devices apply only where the complainant belongs to the protected class, so that the procedural price of complaining is highest for precisely the citizens the statute was enacted to encourage. Parliament was entitled to say that this was not protection against abuse but a transfer of power from complainant to accused at the decisive moment, and it said so in the most direct legislative form available to it. The Court, on review, was then entitled, and bound, to say that its own branch had exceeded its commission: Article 142 empowers the Court to do complete justice in a cause before it, not to legislate for a class of causes, and the distinction between the two is the distinction on which the separation of powers in a written constitution ultimately rests.

The verdict this article reaches is that the recall and the validation are right on the institutional question, and that the episode nevertheless exposes a cost that the tidy ending tends to conceal. On the institutional question there is no serious escape from the review bench’s reasoning. A court that may add an enquiry, an approval and a bail exception to a statute that contains none of them may add anything to any statute, and the power would be limited only by the temperament of the bench exercising it; the same authority invoked in March 2018 to soften the Atrocities Act could be invoked by another bench to soften a statute a reader of this article holds dear, on figures that bench found persuasive. The discipline that interpretation must answer to a text is not pedantry. It is the condition on which judicial review itself is legitimate in a democracy, because it is the only thing that distinguishes a judgment from a vote. The cost, which the tidiness conceals, is that the individual grievance that started the sequence has no institutional home. A public servant against whom a complaint is lodged in retaliation for an official decision, and whose complaint on its face does dress that grievance in the Act’s language, must rely on the threshold jurisdiction preserved in Prathvi Raj Chauhan, on the quashing power, and on the general arrest discipline, all of which operate later, less certainly, and at greater expense than the filter Mahajan tried to build. The honest summary of the sequence is therefore double. The boundary held, and the boundary was worth holding. The problem the boundary prevented the Court from solving remains where the Court found it, and it is, as it always should have been, Parliament’s problem, on which Parliament has given its answer.

One further lesson is methodological, and it may in the long run matter more than the holding. The March 2018 directions rested substantially on a reading of disposal figures: the proportion of cases closed as false and the proportion ending in conviction. The review proceedings exposed how little such figures, unaudited and unexamined, can bear. An acquittal is not a false case; it is a case the prosecution failed to prove, in a field where witnesses are exposed to pressure of exactly the kind the Act exists to counteract, so that a low conviction rate is as consistent with the Act’s necessity as with its abuse. A case closed as false after investigation is a conclusion by the investigating agency whose own conduct the complaint may have implicated. None of this means the figures were fabricated or that abusive complaints do not exist; they do, and the case that began the sequence was one. It means that aggregate disposal statistics cannot distinguish the two explanations, and that a court building general procedure on them is building on ground it has not surveyed. Legislatures make the same error when they legislate from the same figures in the opposite direction, and the discipline this episode teaches both institutions is that the statistics of a contested statute are evidence about the contest as much as about the statute.

The position in practice

For a person accused under the Act, the working position after February 2020 can be stated without ornament. There is no enquiry to insist upon and no approval to await; if the investigating officer considers arrest necessary, arrest may follow registration, subject to the Code’s arrest discipline and to the magistrate’s duty at remand. There is no anticipatory bail where the first information report, read as a whole, discloses the ingredients of an offence under the Act, and no court at the bail stage will weigh evidence or credibility to manufacture an exception. There is anticipatory bail, and there is quashing, where the report on its face discloses no offence under the Act, and the whole of the forensic effort at the threshold goes into that single characterisation question: do the allegations, taken at their highest, satisfy the elements of the clause of Section 3 invoked, including where the clause requires it the public character of the act and the caste-based character of the conduct? The framing of the complaint and the framing of the threshold application are therefore where these cases are won and lost, long before any trial.

The threshold application itself has a settled forensic shape, and it is worth describing because it is so different from an ordinary bail argument. The application is moved before the Court of Session, which for offences under the Act is the Special Court, or before the High Court, and it is argued almost entirely on the text of the first information report. The applicant’s counsel reads the report against the clauses of Section 3 invoked in it, ingredient by ingredient, and the prosecution’s answer is not evidence but the report itself, supplemented at most by the proposition that the ingredients will be supplied by the investigation. Everything therefore turns on drafting, on both sides: on whether the officer who recorded the report captured the statutory ingredients, including, where the clause requires it, the place and its public character and the words alleged to have been used, and on whether the applicant can show a missing ingredient without straying into disputed fact. A report that recites conclusions in the language of the section without alleging the underlying facts is vulnerable; a report that alleges the underlying facts completely is, for threshold purposes, impregnable, however contestable those facts may be. This is a discipline the validated regime imposes on the police as much as on the accused, and it is not the least of the sequence’s ironies that the judgment remembered for removing a police filter ended by making the quality of police drafting decisive at the one screening point that survived.

For a complainant, the position is the mirror image, and it is the position Parliament intended. Registration follows the disclosure of a cognisable offence without a prior police assessment of motive; the investigation proceeds under the Act’s own machinery, including its provisions for the rights of victims and witnesses and for relief; and no official outside the investigating chain holds a power to stop an arrest. What the complainant does not have, and never had, is a guarantee of conviction, and the complainant’s practical vulnerabilities, delay, pressure, and the attrition of a long trial, are addressed by the Act’s special courts and timelines rather than by its procedure at the point of registration. Readers with a professional or examination interest in how this sequence is tested can work through previous-year questions on the amendment and the three judgments in a structured practice viewer, which is the surest way to discover how frequently the two wrong statements about anticipatory bail appear as the attractive wrong options.

For the police officer and the magistrate, the sequence leaves a cleaner instruction set than the one that obtained in the spring of 2018, and it is worth recording what that set is, because the intervening period generated genuine confusion in police stations and remand courts about which regime applied to a complaint lodged on a particular date. The governing instruments are the Act and the Code, and no third instrument. The officer registers, investigates, and arrests where the Code’s conditions for arrest are met, recording reasons as the Code and Arnesh Kumar require. The magistrate authorises detention only on satisfaction recorded in accordance with the Code, examines the legality of the arrest, and entertains the threshold question, where it is raised in the proper proceeding, of whether the complaint discloses an offence under the Act at all. Nothing in that set is peculiar to this statute except the bail exclusion, which is the point the whole sequence was fought to establish.

The account given in this article is a statement of the general legal position across a sequence of judgments and amendments; it is not advice on any individual case, and a person facing proceedings under the Act, or contemplating a complaint, needs advice on the facts of that case from a lawyer who can read the actual complaint, because everything in this field turns on the wording of a single document.

What remains unsettled

Three matters are left open by the validation, and they are structural rather than accidental, so they can be described without any prediction about how or when they will be resolved. The first is the content of the threshold test. Prathvi Raj Chauhan preserves anticipatory bail and quashing where the complaint discloses no offence under the Act, but it does not codify how a court is to treat the recurring borderline patterns: the abuse that was uttered but arguably not in public view, the property or service dispute dressed in caste language, the complaint against a public servant for an act concededly done in an official capacity. High Courts decide these case by case, and their decisions do not all point the same way, so the operative law at the threshold is a body of fact-specific rulings rather than a rule. The second is the interaction between the exclusion and the general law of bail after arrest, in particular how long an accused whose threshold application has failed may expect to remain in custody in a case that later collapses; the Act’s special court timelines bear on this but do not answer it. The third is textual and transitional. Sections 18 and 18A speak of Section 438 of the Code of Criminal Procedure, a code that has been replaced by the Bharatiya Nagarik Suraksha Sanhita, 2023, whose corresponding anticipatory bail provision is Section 482; the exclusion continues to operate through the ordinary rules for construing references to a repealed and re-enacted provision, but the statute book requires the reader to perform that translation, and translations of that kind are a standing invitation to threshold litigation until the text is tidied.

A fourth open matter is institutional, and the mirror sequence is what makes it visible. In the cruelty field, a screening mechanism held in 2018 to be beyond the judicial function returned years later because a High Court had framed it and the Supreme Court adopted it, under its power to do complete justice, without overruling the earlier holding. Nothing in the atrocities recall forecloses an analogous route in form: a High Court remains free to frame supervisory guidelines for the courts and police of its State, and a future bench asked to approve such guidelines in a case before it would have to confront both the recall’s reasoning and Section 18A’s closing words, which forbid any procedure other than that provided under the Act or the Code. Whether that statutory clause distinguishes this field decisively from the cruelty field, where no equivalent clause stood in the way, is a question no decided case has yet had to answer, and it is the point on which the two sequences may yet converge or finally part.

Behind those three sits the question the sequence never answered and could not: whether the figures that launched it, of closure and conviction, reflect a statute that invites abuse or a statute whose cases are exceptionally hard to prove. That question is empirical, it is answerable in principle by research that follows individual cases from complaint to disposal and records why each ended as it did, and neither the Court in 2018 nor Parliament in the same year had such research before it. Until some institution commissions it, argument about the working of the Act will continue to be conducted from the same ambiguous aggregates, and the risk the episode teaches against, of building general procedure on unexamined numbers, will remain live for the next statute and the next bench.

The sequence that began with a store keeper’s grievance about a confidential report therefore ends with an unusually complete map of its own institutional terrain. A court identified a real defect and cured it by a method the Constitution does not allow it. A legislature removed the cure by the method the Constitution prescribes, in words that left no room for doubt about its intention. The same court then withdrew its own work for the right reason, that the method and not merely the outcome was wrong, and finally sustained the legislature’s answer while preserving, at the threshold, the one judicial function that no party had asked it to surrender: the function of asking whether the case before it is a case under the Act at all. A reader who can state those four steps, and say precisely what each decided, understands something about Indian public law that no summary of any single step conveys.

Frequently Asked Questions

Q: What did the Supreme Court direct in March 2018 about the Atrocities Act?

In Dr Subhash Kashinath Mahajan v. State of Maharashtra, decided on 20 March 2018, a two-judge bench held that there is no absolute bar on anticipatory bail where no prima facie case is made out or the complaint is prima facie mala fide; that a public servant could be arrested only with the approval of the appointing authority and any other person only with the approval of the Senior Superintendent of Police, for reasons recorded and scrutinised by the magistrate; and that a preliminary enquiry could be conducted by the Deputy Superintendent of Police before registration, to test whether the allegations made out a case under the Act and were not frivolous or motivated.

Q: Why did Parliament pass the 2018 atrocities amendment?

Parliament passed the Amendment Act of 2018 to remove the three safeguards the Mahajan judgment had read into the Act and to restore the procedure as it stood before March 2018. The government of the day accepted the argument, pressed by organisations of the Scheduled Castes and Scheduled Tribes and by members across parties after widespread protests, that an enquiry before registration and an approval before arrest would delay and deter complaints in precisely the cases the Act was enacted to protect, and that the safeguards were in any event conditions the legislature had never enacted. The amendment therefore inserted Section 18A, negating each direction in terms.

Q: What does Section 18A provide?

Section 18A, inserted by Act 27 of 2018 with effect from 20 August 2018, provides in subsection (1) that a preliminary enquiry shall not be required for registration of a first information report against any person, and that the investigating officer shall not require approval for the arrest, if necessary, of any person accused under the Act, with no procedure applying other than that provided under the Act or the Code of Criminal Procedure. Subsection (2) provides that Section 438 of the Code, the anticipatory bail provision, shall not apply to a case under the Act, notwithstanding any judgment or order or direction of any Court.

Q: Did the Supreme Court recall its own atrocities directions?

Yes. In Union of India v. State of Maharashtra, decided on 1 October 2019 on the Union’s review petition, a three-judge bench of Arun Mishra, M. R. Shah and B. R. Gavai JJ. recalled the arrest approval and preliminary enquiry directions issued in Mahajan, together with the direction that made a breach of them actionable. The bench held that the directions encroached upon the field reserved for the legislature, were inconsistent with the concept of protective discrimination under Article 15(4) of the Constitution, and were an impermissible use of the Court’s power under Article 142, which cannot be used to supplant a statutory scheme.

Q: Was Section 18A upheld as constitutional?

Yes. In Prathvi Raj Chauhan v. Union of India, decided on 10 February 2020, a three-judge bench upheld Section 18A against a challenge founded on Articles 14 and 21 and on the separation of powers. The Court reasoned that Parliament, in inserting the section, had changed the underlying law within its legislative competence in response to a judgment, which a legislature is entitled to do, and that the stringency of the procedure served the Act’s protective object. The Court added that the provisions had in large measure been rendered academic by the recall of the Mahajan directions in the review decided the previous October.

Q: Can anticipatory bail ever be granted in an atrocities case now?

Yes, but only on one ground and within narrow limits. The bar in Sections 18 and 18A operates where the complaint discloses an offence under the Act. Where the complaint, on its face, does not disclose the ingredients of an offence under the Act at all, there is no case under the Act for the bar to attach to, and the ordinary jurisdiction of the courts remains, so anticipatory bail may be granted. A court deciding the question reads the complaint; it does not weigh evidence or credibility, and it will not grant bail merely because the allegation appears improbable or the arrest would be harsh.

Q: What does the atrocities amendment sequence show about judicial and legislative roles?

It shows the boundary being contested and then settled from both directions. A court read procedural safeguards into a complete statute; the legislature removed them by amending the statute in general terms; the same court then recalled its own directions as judicial legislation, on grounds independent of the amendment; and a bench of the court upheld the amendment with a clarification of its own. The sequence demonstrates that a legislature may answer a judgment by changing the law, that it may not merely annul the judgment, and that a court’s power to do complete justice in a case does not extend to enacting procedure for a class of cases.

Q: What happened during the protests against the atrocities ruling?

In the weeks after the judgment of 20 March 2018, organisations of the Scheduled Castes and Scheduled Tribes protested across the country against what they characterised as a dilution of the Act, and the protests culminated in a nationwide day of protest at the beginning of April 2018 during which lives were lost. The political consequence was immediate: the Union of India sought review of the judgment and committed itself to legislation, and Parliament enacted the amendment in August 2018. The protests are recorded here as part of the legislative history; responsibility for the deaths that occurred is a contested matter on which this article makes no finding.

Q: What was the Mahajan case about?

The case began as an appeal by Dr Subhash Kashinath Mahajan, Director of Technical Education in Maharashtra, against proceedings under the Act lodged against him by a subordinate employee. The employee had complained against two senior officers who made adverse entries in his confidential report; sanction to prosecute those officers was sought from the appellant and refused; the employee then filed a first information report against the appellant himself, alleging that he had not been competent to decide the sanction question. The Supreme Court quashed the proceedings as an abuse of process, but used the case to frame general questions about safeguards against abusive invocation of the Act, which produced the directions of March 2018.

Q: What did the October 2019 review judgment hold?

The review judgment in Union of India v. State of Maharashtra held that the operative directions in Mahajan, the arrest approvals and the preliminary enquiry, deserved to be recalled, and it recalled them. It reasoned that the directions added conditions to a statute the legislature had enacted without them, created a procedural regime that burdened complainants under a protective statute alone, conflicted with the Constitution Bench position in Lalita Kumari that registration follows disclosure of a cognisable offence, and exceeded the limits of the power under Article 142 as explained in the Supreme Court Bar Association case. The quashing of the proceedings in Mahajan itself was not the subject of the recall.

Q: What did Prathvi Raj Chauhan decide in February 2020?

Prathvi Raj Chauhan v. Union of India, decided on 10 February 2020, upheld the constitutional validity of Section 18A and of the exclusion of anticipatory bail in cases under the Act. The bench of Arun Mishra, Vineet Saran and S. Ravindra Bhat JJ. held that Parliament had permissibly changed the law in response to the Mahajan directions, and clarified that where a complaint on its face does not disclose an offence under the Act, the bar on anticipatory bail does not apply and the ordinary jurisdiction of the courts, including the power to quash, remains available. Justice Bhat, concurring, stressed that anticipatory bail in such cases is exceptional.

Q: What is the difference between Section 18 and Section 18A of the Act?

Section 18 is part of the original Act of 1989 and does one thing: it provides that nothing in Section 438 of the Code of Criminal Procedure shall apply in relation to any case involving the arrest of a person on an accusation of an offence under the Act. Section 18A, inserted in 2018, is a response to the Mahajan judgment. It adds the two procedural negatives, no preliminary enquiry before registration and no approval before arrest, with the direction that no procedure other than that provided under the Act or the Code shall apply, and it repeats the Section 438 exclusion with the words “notwithstanding any judgment or order or direction of any Court” added.

Q: Why did the review bench describe the 2018 directions as judicial legislation?

Because the directions did not interpret any provision of the Act. No party had contended that Section 18 was ambiguous, and the enquiry and approval machinery had no anchor in the Act’s text; the bench had created designated officers, recorded reasons, magisterial scrutiny and sanctions for breach, which is the characteristic content of legislation, and had done so for a statute Parliament had comprehensively revised two years earlier without including any of it. On the review bench’s reasoning, drawn from the limits the Court has placed on its Article 142 power, a court may do complete justice in the case before it but may not build a procedural edifice the legislature has declined to build.

Q: What material did the March 2018 bench rely on when it read in safeguards?

The bench relied principally on the facts of the case before it, which it found to be a clear abuse of process, and on figures of the National Crime Records Bureau concerning cases under the Act, which it read as showing that a substantial proportion of investigated cases were closed as false and that convictions formed a small proportion of disposals. From that material it inferred what it called the acknowledged abuse of the law of arrest under the Act. In the review proceedings the Union contested the inference, arguing that acquittals and closures record the outcomes of investigations and trials and do not establish that complaints were false, an evidential dispute the review bench noted without needing to resolve.

Q: Does any preliminary enquiry survive before an FIR under the Act?

Not as a step in the procedure of the Act. Section 18A provides that a preliminary enquiry shall not be required for registration of a first information report against any person, and the Mahajan direction for a Deputy Superintendent’s enquiry has been recalled. What survives is the general position in Lalita Kumari v. Government of Uttar Pradesh, where a Constitution Bench held registration mandatory on disclosure of a cognisable offence and permitted a preliminary enquiry only in defined categories of case, such as matrimonial and family disputes, commercial offences, medical negligence and corruption, to determine whether a cognisable offence is disclosed at all. Atrocity complaints are not within those categories, and the validation judgment treated the enquiry question on that footing.

Q: How does the atrocities sequence compare with the Section 498A sequence?

In both fields a two-judge bench created a screening mechanism the statute did not contain: district Family Welfare Committees for cruelty complaints in Rajesh Sharma in 2017, and enquiry and approval requirements for atrocity complaints in Mahajan in 2018. The corrections differed. In the cruelty field a three-judge bench set the committee directions aside in Social Action Forum for Manav Adhikar in 2018 as beyond the judicial function, and Parliament never legislated; in the atrocities field Parliament legislated first, in August 2018, and the Court recalled its directions afterwards. The cruelty mechanism later returned through guidelines framed by a High Court and approved by the Supreme Court in 2025, a route the atrocities sequence has not taken.

Q: What role did Article 142 of the Constitution play in the recall?

Article 142, which empowers the Supreme Court to pass such orders as are necessary for doing complete justice in any cause or matter pending before it, was the ground on which the recall principally rested. The Mahajan directions had been defended as an exercise of the Court’s power to secure a just and fair procedure; the review bench held, applying the limits stated in Supreme Court Bar Association v. Union of India, that the power operates within the law, supplements it where it is silent, and cannot be used to supplant a substantive statutory scheme or to enact conditions the legislature has omitted. Because the directions failed that test, they were recalled as an impermissible use of the power.

Q: What did Justice S. Ravindra Bhat say in his concurring opinion in 2020?

In Prathvi Raj Chauhan, Justice S. Ravindra Bhat concurred in upholding Section 18A and wrote separately to emphasise two things. The first is constitutional value: the Preamble’s fraternity and the promise of equal citizenship are, in his reasoning, considerations of the same order as the personal liberty of an accused person, because the Act exists to make that citizenship real for communities historically denied it. The second is discipline at the threshold: anticipatory bail in a case under the Act is exceptional and sparing, available where the complaint does not disclose a prima facie case under the Act, and the threshold enquiry must not be allowed to grow into an assessment of the evidence or of the complainant’s credibility.

Q: What did the Union of India argue in the review petition?

The Union of India, which administers the Act through the Ministry of Social Justice and Empowerment, argued that the Mahajan directions had rewritten a protective statute under the guise of interpreting it; that they rested on an impermissible use of Article 142; that they discriminated in effect against complainants from the Scheduled Castes and Scheduled Tribes by subjecting their complaints, alone among complaints of cognisable offences, to a prior enquiry and an arrest approval; and that the disposal figures relied on in the judgment could not establish that complaints were false. The Union also defended, in the parallel challenge proceedings, Parliament’s competence to enact Section 18A as a change in the underlying law.

Q: Why does Section 18A say “notwithstanding any judgment or order or direction of any court”?

Because the provision was drafted to displace a specific judicial act. Section 18 had excluded anticipatory bail since 1989, but the Mahajan judgment had held the exclusion to be qualified and had issued directions inconsistent with it. The notwithstanding clause in Section 18A(2) instructs every court that the exclusion operates whatever any court has said or directed, removing the precedential basis on which a subordinate court might have continued to apply the directions. Its legal effect is prospective and general: it changes the law for future cases rather than annulling the Mahajan judgment between its parties, which is why the clause was upheld as legislation rather than condemned as an attempt to reverse a decision.