The Shreya Singhal judgment answered a question the text of the Information Technology Act, 2000 had left dangerously open: how far may Parliament go in criminalising speech sent through a computer, when the Constitution permits restrictions on expression only on eight enumerated grounds and for no others? On 24 March 2015, a bench of the Supreme Court composed of Chelameswar J. and Nariman J., with the judgment delivered by Nariman J., held that one provision had crossed every line the Constitution draws, that a second provision survived because it stayed within those lines, and that a third survived only after the Court rewrote its operative words by construction. The case, reported as Shreya Singhal v. Union of India, (2015) 5 SCC 1, is the leading Indian authority on speech on the internet, and the doctrinal part of it can be stated in a sentence: discussion and advocacy, however unpopular, sit at the heart of the freedom in Article 19(1)(a), and only incitement may be reached by a restriction under Article 19(2).

What gives the judgment its second life, and what makes it the sharpest case study in this series on the distance between a judgment and its effect, is what happened afterwards. The provision the Court destroyed did not disappear. It remained printed in the statute book, annotated with a footnote, and police stations and trial courts across the country went on using it. Research built on district court records, conducted by the Internet Freedom Foundation with CivicDataLab and placed before the Court by the People’s Union for Civil Liberties, recorded that in the eleven States the research covered, 229 cases under the provision were pending when the judgment was delivered, and 1,307 fresh cases were lodged under it after it had ceased to exist in law. A provision that was void in law therefore produced more prosecutions in the six years after the judgment than were pending at the moment it fell. This article reads the judgment as the Court wrote it, holding by holding, and then reads the record of its implementation with the same care, because the second reading explains a structural feature of Indian law that the first reading alone would never reveal. The wider field in which the case sits, and the map of the statutes and rules that govern online content, is set out in the series guide to technology law in India, to which this article is the judicial companion.

The Provisions in Issue and the Ambiguity in the Text

Four provisions stood before the Court. The first, and the one on which everything turned, was Section 66A of the Information Technology Act, inserted by the amendment of 2008 and in force from 27 October 2009. The provision’s own passage story, the debate that did not happen and the drafting choices that produced its wording, belongs to the article on the 2008 amendment that inserted Section 66A, and is taken here only as far as the Court took it. As the judgment itself noted, in a footnote tracing the genealogy of the language, the provision’s ancestry ran back to Section 10(2)(a) of the United Kingdom’s Post Office (Amendment) Act, 1935, which penalised sending a grossly offensive, indecent, obscene or menacing message by telephone, carried forward in Section 66 of the Post Office Act, 1953. A form of words written for the telephone network of the 1930s, where a message passed between identified persons, had been transplanted in 2008 into a medium where a message could be published to the world, and the transplant carried none of the limiting context of its origin.

Section 66A punished any person who sent, by means of a computer resource or a communication device, information that was grossly offensive or had a menacing character. It punished information which the sender knew to be false, sent persistently for the purpose of causing annoyance, inconvenience, danger, obstruction, insult, injury, criminal intimidation, enmity, hatred or ill will. And it punished an electronic mail or message sent for the purpose of causing annoyance or inconvenience, or to deceive or mislead the recipient about its origin. The punishment on conviction was imprisonment for a term which could extend to three years, together with a fine. The provision made no distinction between a private message and a public post, between a communication to one person and a publication to millions, between speech addressed to an individual and speech addressed to the public on a public question. It required, in its first limb, no falsity, no intent to harm and no consequence at all: it was enough that the information be grossly offensive or menacing in character, judged by a standard the statute did not supply.

What kinds of messages did Section 66A cover?

Section 66A covered three kinds of sending: information that was grossly offensive or menacing in character; information known to be false and sent persistently to cause annoyance, inconvenience, danger, obstruction, insult, injury, criminal intimidation, enmity, hatred or ill will; and email sent to annoy, inconvenience, deceive or mislead the recipient about its origin.

Nothing in the text confined the offence to threats, fraud or harassment of an identified person, and nothing required that anyone in fact be annoyed, endangered or misled.

The second provision was Section 69A, which empowers the Central Government, where satisfied that it is necessary or expedient so to do in the interest of the sovereignty and integrity of India, the defence of India, the security of the State, friendly relations with foreign States, public order, or for preventing incitement to the commission of a cognizable offence, to direct that access by the public to information through a computer resource be blocked. The procedure and safeguards were supplied by the Information Technology (Procedure and Safeguards for Blocking for Access of Information by Public) Rules, 2009, which require reasons to be recorded and route a blocking request through a committee, with a review mechanism above it. The third provision was Section 79, the intermediary immunity clause, under which an intermediary is not liable for third party information, data or a communication link made available or hosted by it, subject to conditions. Section 79(3)(b) withdrew that immunity where, upon receiving actual knowledge, or on being notified by the appropriate government or its agency, that the material was being used to commit an unlawful act, the intermediary failed expeditiously to remove or disable access. Rule 3(4) of the Information Technology (Intermediaries Guidelines) Rules, 2011 imposed a parallel duty: an intermediary, on obtaining knowledge by itself or on being brought to actual knowledge by an affected person in writing or through email, was required to act within thirty six hours to disable information falling within the prohibited categories listed in Rule 3(2). The fourth provision, Section 118(d) of the Kerala Police Act, 2011, punished causing annoyance to any person, or a nuisance, through a communication device or by any other means, in a manner the provision left similarly open.

The anatomy of the first provision deserves a slower reading, because the judgment’s conclusion is unintelligible without it. Section 66A had three limbs, and each limb failed in a different way. The first limb punished the sending of information that is grossly offensive or has a menacing character. These are judgements of taste and apprehension dressed as elements of an offence. Grossly offensive measures the reaction of a reader against a standard that the statute does not state and that no cited authority had ever fixed, and menacing character asks whether information, in itself, wears a threatening aspect, without requiring any threat to be conveyed to any person or any intent that it be taken as one. The second limb punished information known to be false, sent persistently, for the purpose of causing a list of consequences that runs from annoyance through to criminal intimidation. The list is the provision’s confession of its own method. Annoyance and inconvenience are the small change of daily communication, insult and ill will are the ordinary currency of political argument, and danger, obstruction, injury and criminal intimidation are grave matters of a different order; a text that places them in one undifferentiated sequence, governed by one punishment, is not defining an offence but forbidding a climate. The requirement of persistence and of knowledge of falsity, the two elements that might have disciplined the limb, attached only to part of it and disciplined nothing else. The third limb punished email sent to annoy or inconvenience, or to deceive or mislead the recipient about its origin, and so carried the same standard down to the most routine act of electronic life. Across all three limbs one structural feature recurs, and the judgment fastened on it: the offence was complete on sending. The statute did not require that the message reach its target, that the target be affected, that any public be disturbed, or that the sender intend the prohibited character the message was later found to have. A provision of that design does not regulate harm. It licenses the state, at the instance of any complainant who professes the listed reactions, to punish expression as such.

The remaining provisions had anatomies of the opposite character, and the difference in design is what the judgment ultimately vindicated. Section 69A does not create an offence at all. It is a power of the executive, bounded by its text to the grounds that Article 19(2) itself enumerates, exercisable by a direction to block access, and surrounded by the Blocking Rules with a documentary procedure: a request in a prescribed form, examination by a committee constituted under the Rules, an opportunity to the person who placed the information and to the intermediary to be heard where they can be identified, the recording of reasons, and periodic review by a review committee constituted under the telegraph framework. Whatever the criticisms of that machinery, and they are taken up in full below, the power comes with its standards printed on its face and its grounds closed by enumeration. Section 79 is a provision of a third kind, an immunity with conditions, and its context matters to the reading the Court gave it. The immunity is the commercial foundation of the platform economy: without it, every host, access provider and marketplace would carry liability for the speech of every user, and the internet as a medium of publication by non professional speakers could not operate. The conditions in Section 79(2), that the intermediary’s function be limited to providing access, that it not initiate, select or modify the transmission, and that it observe due diligence and the guidelines, locate the intermediary as a conduit. Section 79(3)(b) then withdrew the immunity on actual knowledge and failure to act, and Rule 3(4) translated that withdrawal into a thirty six hour takedown duty triggered by a private complaint. The constitutional difficulty in that arrangement is not visible until its consequences are traced: the trigger was a complaint, the adjudicator was the company, and the penalty for adjudicating wrongly was liability for other people’s speech. The fourth text, Section 118(d) of the Kerala Police Act, returned to the pattern of the first: a police power framed around the causing of annoyance, in a state Act of general application, penal in operation and standardless in terms.

It follows that the four challenges, although argued together and governed by one set of principles, presented the Court with four distinguishable juridical objects: an offence without a standard, a power with enumerated grounds and a documented procedure, an immunity with an adjudicatory trigger, and an offence without a standard in a second legislature’s hand. The judgment is remembered as a free speech landmark, which it is, but its method is that of a court reading texts one at a time and letting the results differ where the texts differ. That method also explains the limits of the victory the petitioners won. Nothing in the reasoning casts doubt on the state’s power to punish threats, fraud, impersonation or incitement under provisions that define them, or to block access on the constitutional grounds by the documented route, or to fix intermediaries with liability on adjudicated knowledge. What the Court removed from the statute book was a technique, the technique of criminalising expression by the reactions it might provoke, and the afterlife of the case is the story of how long a removed technique can persist when its words remain in print.

The ambiguity the Court had to resolve was not hidden in any corner of these texts. It stood on their face. Words such as grossly offensive, menacing, annoyance and inconvenience have no fixed legal content. They describe reactions, and reactions vary with the person reacting. A provision that punishes a sender by reference to the predicted reaction of an audience the sender cannot identify, in a medium where the audience is everyone, asks the citizen to guess at the boundary of a criminal offence before speaking. The constitutional question was whether such guessing is a price the freedom in Article 19(1)(a) can be made to bear, and that question could only be answered by measuring the provision against the closed list in Article 19(2): the sovereignty and integrity of India, the security of the State, friendly relations with foreign States, public order, decency or morality, contempt of court, defamation, and incitement to an offence. A restriction on speech that cannot be brought under one of those heads is not a reasonable restriction at all. It is a prohibition the Constitution does not authorise, and the whole judgment is an application of that discipline to four different texts, with four different outcomes.

The Road to the Court

The challenge reached the Court on the strength of arrests that the country read about. In November 2012, two young women, Shaheen Dhada and Renu Srinivasan, were arrested in Palghar, in Thane district of Maharashtra, after one of them posted a comment on a social networking site questioning the shutdown of Mumbai on the death of the Shiv Sena leader Bal Thackeray, and the other indicated approval of the post. Section 66A supplied the arresting power. The arrests produced public anger of a kind that attaches to cases in which the distance between the words used and the force used against them is visible to everyone, and a young law student, Shreya Singhal, filed a writ petition under Article 32 asking that the provision be struck down. Her petition, Writ Petition (Criminal) No. 167 of 2012, became the lead case in a batch of ten writ petitions heard and decided together. The batch included a petition by the cartoonist Aseem Trivedi, who had himself faced arrest, a petition by the consumer review platform Mouthshut.com, which lived under the takedown regime as an intermediary, and petitions by the People’s Union for Civil Liberties, by Common Cause, and by the Internet and Mobile Association of India, whose members carried the intermediary obligations the case would construe. The petitioners attacked Section 66A principally under Article 19(1)(a), with arguments under Articles 14 and 21 in support, and attacked Sections 69A and 79 and the Kerala provision on related grounds. The provision’s insertion history, including the speed of its passage in December 2008, is treated in the amendment article linked above and is not repeated here, because the Court’s analysis did not depend on it.

The executive had, by the time of the hearing, already conceded in practice that the provision was being used in ways its text invited. In January 2013, in the weeks following the Palghar arrests and the criticism they had drawn from every quarter, the Central Government issued an advisory under which no arrest was to be made under Section 66A without the approval of a senior police officer, an inspector general in metropolitan areas and a deputy commissioner or superintendent elsewhere. The advisory is part of the record of this case because it exposed the vice the Court would later name: if a provision requires an administrative filter between its text and its use, the filter is an admission that the text itself does not tell the officer at the police station where the offence begins. The judgment would return to this point in doctrinal form, in its treatment of the argument that a law cannot fall merely because it is capable of abuse.

What did the Union government argue in defence of Section 66A?

The Union argued that the provision was aimed at a harm, the misuse of a medium of great reach and speed, and that its words had to be read loosely because the mischief was various. It argued that possible abuse is no ground to strike down a law, and that the Court should presume the administering authority would act reasonably.

It also undertook that the provision would be administered with restraint, and pointed to the advisory requiring senior approval before arrest as evidence of that restraint. The loose language, it said, was necessary because Parliament could not enumerate every form that online abuse might take, and the provision’s breadth was the proportionate response of a legislature confronting a medium whose abuses change shape faster than statutes can be amended.

The petitioners’ answer, which the judgment in substance accepted, was that the argument proved too much. A criminal statute addressed to the public at large cannot be saved by an assurance about how it will be administered, because the citizen deciding whether to speak has only the text to go by, and the text is addressed to her, not to the administrator. The petitions were heard over several days and the judgment, reserved and then delivered on 24 March 2015, ran to a full treatment of the free speech jurisprudence of the Court from its earliest cases forward. Its structure matters for understanding its authority. The Court first settled the governing principles, then tested Section 66A against them head by head under Article 19(2), then applied the doctrines of vagueness and overbreadth, then considered and refused the request to save the provision by narrowing it, and only then turned to the blocking power, the intermediary provisions and the Kerala provision, in each of which the analysis produced a different result because the texts were different. The judgment is therefore not one holding but a controlled experiment in constitutional method: the same principles, applied to four provisions, striking two down, upholding one, and construing one into validity.

The Questions the Court Framed for Itself

Constitutional judgments of this weight are usually organised around a small number of questions, and the shape of the questions explains the shape of the answers. The first question was one of fit. Article 19(2) is a closed list, and the Court treated it as closed in both directions: a restriction must find its warrant in one of the eight heads, and the warrant must cover the actual reach of the provision, not merely some application of it. That framing disposed, in advance, of the most common style of defence offered for speech offences, which is to point to the worst conduct the provision might catch and to ask that the provision be judged by it. A law that punishes the grossly offensive is not saved because some grossly offensive material is also a criminal intimidation; it must be judged by the whole of what it makes punishable, and the heads must cover that whole, severally and actually.

The second question was one of standard. The Court asked what performance the Constitution demands of the language of a speech offence. The answer it assembled, from its own earlier cases and from the comparative authorities, is that such language must satisfy two audiences at once. It must tell the citizen, in advance and with reasonable definiteness, what is forbidden, so that speech is regulated by the speaker’s own assessment and not by the speaker’s fear. And it must confine the enforcing officer, by boundaries that a court can police, so that the decision to prosecute does not become a discretion to select among speakers. A term that fails the first audience produces the chilling effect; a term that fails the second produces arbitrary enforcement; Section 66A was attacked, and fell, on both. The judgment’s repeated insistence that the expressions used are undefined is therefore not a complaint about draftsmanship in the abstract. It is the finding that neither audience had been given anything to work with.

The third question was one of remedy, and it structured the whole second half of the case. Faced with an unconstitutional text, a court may strike it down, sever the invalid part, or read it down into validity, and the petitioners and the Union argued all three across the four provisions. The Court’s handling of the remedial menu is the analytical core of the judgment, and it is rigorous in a way commentary sometimes misses. Reading down is permissible where the text has a settled or natural meaning within constitutional bounds that the court can adopt as the provision’s true construction; it is not permissible where the court would have to supply every operative element itself. Severance is permissible where what remains is a coherent provision that Parliament can be taken to have enacted independently; it is not permissible where the invalidity is the provision’s essence. Applying that remedial discipline, the Court read Section 79 into validity because a construction was available that the text would bear, upheld Section 69A because no remedial question arose, and refused both techniques for Section 66A and Section 118(d) because in each case the vice was the offence itself. The remedial analysis, more than any single proposition about speech, is what a legislature must study before drafting in this field, because it tells the drafter that the court will neither rescue a standardless text nor refuse a salvageable one.

There is a unity to the batch that a reader of the four holdings in isolation can miss, and the Court’s willingness to decide the offence, the blocking power and the immunity in one judgment is what discloses it. The three provisions are the three points at which the state touches online speech. It touches the speaker, by the criminal law. It touches the channel, by the blocking power. And it touches the host, by the conditions of immunity. A defect at any one point distorts the others: a vague offence makes hosts cautious regardless of their immunity, since hosting is no comfort to a user in custody; a complaint driven takedown regime makes the criminal law’s limits academic, since speech can be extinguished without any offence being charged; and an unreviewable blocking power would let the executive achieve by direction what the offence could not achieve by prosecution. The judgment can be read, and is best read, as a single survey of that triangle, in which the Court asked of each point the same question in the form appropriate to it: where does the decision that speech is unlawful get made, by whom, under what standard, and subject to what check? The offence failed because the decision was made by the complainant and the constable under no standard. The takedown trigger failed in its unamended form because the decision was made by the host under a private standard. The blocking power survived because its decision is made by a named public authority under enumerated grounds, with reasons, a committee and a court behind it. Stated at that level of generality, the judgment is a decision about the constitutional allocation of the power to judge speech, and its holdings on the three provisions are one holding in three applications. That is also why the case has organised the field since: every subsequent proposal for the regulation of online content in India, whether by statute, by rule or by direction, has had to answer the judgment’s question before it answers any other, which is not what the proposal does to speech but who it appoints to decide.

Two further framing choices deserve record. The Court entertained the challenges under Article 32 at the instance of speakers, platforms and public interest organisations together, and the presence of the platforms in the batch mattered: the intermediary questions could be argued by the parties who would have to operate the machinery the judgment was construing, and the takedown analysis is correspondingly practical, concerned with volumes of complaints and the economics of removal rather than with abstractions. And the Court framed the stakes in the language of the citizen rather than of the publisher. The judgment’s free speech reasoning repeatedly returns to the position of the ordinary user of the medium, the person whose post, comment or cartoon is the typical object of the provision, and it contains the passage, much quoted since, that the public’s right to know is directly affected by Section 66A. That passage is the bridge between the two halves of the case. The right to know belongs to the audience, and a provision that prices speakers out of speaking injures an audience that never learns what it has lost, which is why a court cannot measure such a provision only by the prosecutions under it: its largest effect is the speech that never occurs.

What the Shreya Singhal Judgment Held

The judgment’s conclusions are collected in its final operative paragraph, and they repay being kept separate, because commentary on the case routinely merges them. Section 66A was struck down in its entirety as violative of Article 19(1)(a) and not saved under Article 19(2). Section 69A and the Blocking Rules of 2009 were held constitutionally valid. Section 79 was held valid subject to Section 79(3)(b) being read down, and the Intermediaries Guidelines Rules of 2011 were held valid subject to Rule 3(4) being read down in the same manner. Section 118(d) of the Kerala Police Act was struck down as violative of Article 19(1)(a) and not saved under Article 19(2). Each conclusion rests on reasoning the article takes in turn, beginning with the principle that controls the first of them.

Before the holdings, a word about the judgment’s own account of what was at stake, because its register explains its later treatment by every court that cites it. The Court did not treat the case as a dispute about the management of a troublesome technology. It treated it as a case about the conditions of democratic government, in which the freedom to discuss public questions, including foolishly, offensively and in bad taste, is not a concession the state makes to the citizen but a structural feature of the only kind of state the Constitution establishes. The reasoning moves throughout between two registers that Indian speech cases often keep apart: the register of doctrine, in which heads, proximities and standards are tested, and the register of consequence, in which the Court attends to the behaviour a provision produces in speakers who will never litigate, in officers who will never read the judgment, and in audiences who will never know what they were not told. The presence of the second register is what makes the judgment’s afterlife so pointed an irony. This was a court unusually conscious, on the face of its reasons, that a speech law operates through anticipation and deterrence far beyond its prosecutions; and the one audience whose behaviour the judgment did not anticipate was the state’s own, whose anticipation of nothing, not deterrence, not instruction, not even curiosity about a footnote, allowed the condemned provision to outwork its replacement in the only currency the provision dealt in, which is cases.

The controlling principle is stated early in the judgment and is the proposition for which the case is most often cited. The Court held that three concepts recur through the free speech cases and must be kept distinct: discussion, advocacy and incitement. Discussion or advocacy of a particular cause, no matter how unpopular, is at the heart of Article 19(1)(a). It is only when such discussion or advocacy reaches the level of incitement that Article 19(2) becomes available to restrict it, and a law restricting speech must then fall within one of the specified grounds and be proximate to the harm the ground names. The distinction is not verbal decoration. It fixes the point at which the state’s power begins, and it locates that point at the far end of the spectrum of speech, at the moment speech stops being an appeal to the judgment of others and becomes a trigger for the unlawful act itself. Everything else in the speech part of the judgment is machinery for enforcing that location: the insistence on a proximate connection between the speech and the apprehended harm, drawn from the Court’s earlier public order cases, including the requirement in S. Rangarajan v. P. Jagjivan Ram that the connection be direct, like a spark in a powder keg, and not remote or conjectural; the refusal to let a provision draft its restriction wider than the ground it invokes; and the holding that a restriction which fails the grounds cannot be rescued by the importance of the medium or the sincerity of the legislature.

Measured against the grounds, Section 66A failed at the threshold and at every point after it. The Court took the heads of Article 19(2) one by one. A very large part of what the provision covered, information that is grossly offensive, or that causes annoyance, inconvenience, insult or ill will, answered to no head at all: offensiveness is not public order, annoyance is not decency or morality, and insult is not defamation, which has its own law and its own defences. Where the provision touched the territory of a head, it overshot it. Its public order limb required no tendency to disturb public order at all, proximate or otherwise, and so reached discussion and advocacy that the Constitution protects; its incitement limb was not confined to incitement, because the provision punished the causing of enmity, hatred and ill will as communication outcomes rather than as instigation to an act; and material that fell within it could be, and demonstrably had been, entirely innocent speech on public questions, from a comment on a city shutdown to cartoons about corruption. The Court put the conclusion in terms of coverage: the provision created an offence against persons who use the internet to disseminate information, it took into its net a very large amount of protected and innocent speech, and it arbitrarily, excessively and disproportionately invaded the right of free speech, upsetting the balance between the right and the restrictions that Article 19(2) permits. A provision whose ordinary operation is to catch what the Constitution protects cannot be described as a reasonable restriction on a specified ground, and no redrafting of the analysis under some other head could change the coverage that produced the failure.

The public order head repays separate treatment, because it was the head under which the Union’s case was strongest in appearance and weakest in law. The Court’s own jurisprudence on public order is a jurisprudence of proximity. In the cases following the Constitution’s commencement, the Court had read the head narrowly, and the First Amendment’s addition of public order to Article 19(2) had been accompanied by the discipline that the restriction must have a proximate connection with the disorder apprehended, a connection the Court in Superintendent, Central Prison v. Ram Manohar Lohia required to be direct rather than remote, fanciful or far fetched. In S. Rangarajan v. P. Jagjivan Ram, a case about a film, the Court had expressed the same demand in the image of the spark in a powder keg: the expression must be inseparably locked up with the disorder it is said to threaten. Section 66A contained no element that performed this work. It did not require that the information tend to disturb public order, that it be addressed to a crowd, that it occur in circumstances of tension, or that any disorder be apprehended at all. The provision’s career was the demonstration of the omission. The prosecutions that had brought the case to court, and the further instances pressed in argument, were prosecutions for a comment on a public shutdown, for cartoons about corruption and Parliament, for criticism of political leaders, for speech, in short, at the centre of the democratic process, in circumstances where no disorder beyond the anger of the persons criticised was alleged. A provision defended under public order that operates, in its documented applications, against commentary on public events has failed the proximity test in fact as well as in law, and the judgment’s treatment of the head consists largely in holding those two findings together.

The decency or morality head failed for a cognate reason. The concepts the head protects are constitutional concepts with their own case law, and a message may be grossly offensive to a section of opinion without approaching the territory the head occupies, which is concerned with obscenity and kindred categories the law has spent decades defining. The defamation head failed for a different reason, which exposes the provision’s indifference to legal structure: defamation is a defined wrong with a defined subject, the reputation of an identifiable person, and defined defences, of which truth and fair comment are the chief, and Section 66A’s vocabulary of insult and ill will borrows the emotional register of defamation while discarding every one of its safeguards, punishing the insult without requiring falsity, a victim or the loss of a defence. A legislature that wishes to protect reputation has a law for the purpose, and the existence of that law, with its built in balance between reputation and speech, is the measure of what Section 66A omitted. The incitement head was the only one on which the provision and the Constitution share a word family, and even there the fit failed, because the provision punished the causing of enmity, hatred and ill will as states of mind produced in an audience, where the head permits restriction of incitement to an offence, an act directed at producing unlawful conduct. Between stirring ill will and inciting an offence lies the whole territory of heated political speech, and the provision stood squarely on the wrong side of it.

The second line of reasoning was vagueness, and it is here that the judgment did its most durable doctrinal work. The Court reviewed the authorities on void for vagueness, including the American cases from which the doctrine comes and the Indian decisions, among them Kartar Singh v. State of Punjab, that had accepted vagueness as a ground of challenge where a penal law fails to define the offence with sufficient definiteness that ordinary people understand what conduct is prohibited and in a manner that does not encourage arbitrary and discriminatory enforcement. Section 66A failed that standard in a way the Court treated as incurable by interpretation. None of its operative expressions is defined in the Act. Grossly offensive, menacing, annoyance, inconvenience, danger, obstruction, insult, injury, criminal intimidation, enmity, hatred and ill will do not merely sit at the edge of a defined core; they are the whole content of the offence, and they point in different directions, some to the character of the information, some to the purpose of the sender, some to the feelings of the recipient. What is offensive to one person is not offensive to another, and a sender on a public network cannot know in advance which of the world’s possible recipients will supply the reaction that completes the offence. The Court also treated the provision as overbroad in the related sense: it swept in speech that no permissible restriction could reach, and its breadth was not incidental but structural, since the provision policed the ordinary vocabulary of online disagreement. The practical consequence the Court named was the chilling effect. A speaker who cannot know in advance whether a sentence is an offence will insure against the risk by silence, and the speech lost is precisely the discussion and advocacy the first part of the judgment placed at the centre of the freedom. The provision thus damaged the right twice over, once by punishing protected speech and once by deterring speech that would never have been prosecuted at all.

Why did the Court refuse to save Section 66A by reading it down?

Because there was nothing to read down to. A court may save a provision by confining it to the constitutional core its words will bear, as the sedition provision had been saved in Kedar Nath Singh by confining it to incitement to violence or public disorder. Section 66A had no such core.

Its words did not overshoot a valid offence at the margins; the offence itself was made of the overshoot, and to confine it the Court would have had to write a different provision, choosing the threshold, the mental element and the harm afresh. That is legislation, and the Court declined it.

The severance argument failed for the same reason. Severance removes an invalid part and leaves a valid remainder, but every limb of Section 66A drew on the same undefined vocabulary and the same missing connection to Article 19(2), so there was no remainder that could stand. The argument that a statute cannot be invalidated for the possibility of abuse, pressed by the Union on the strength of the Court’s older cases, was answered by distinguishing them: the vice here was not that a valid power might be abused, but that the power conferred was itself the power to punish protected speech, so the abuse was the ordinary working of the provision and not a departure from it. The assurance of restrained administration met the same answer. A law that is unconstitutional in its terms is not saved by the good intentions of the government of the day, because governments change and the text remains; the Court would later have bitter reason, in the implementation proceedings described below, to reflect on how completely its reasoning assumed that a declaration of invalidity reaches the ground. On the challenge under Article 14 to the singling out of online speech, the Court accepted that the legislature could treat the internet as a distinct field of regulation given the medium’s reach and permanence, and the classification argument failed; the provision fell not because it addressed the internet but because what it did there could not be brought within the only clause that permits speech to be restricted.

Section 69A produced the opposite result, and the contrast is the clearest proof that the judgment is a judgment about texts and not about outcomes. The blocking power was upheld because it is, in the Court’s reading, narrowly drawn. Its grounds reproduce the Article 19(2) heads and add nothing beyond them, so the threshold failure that destroyed Section 66A does not arise. The power is exercisable on the government’s satisfaction that blocking is necessary or expedient on those grounds, the reasons must be recorded in writing, and the Blocking Rules of 2009 surround the direction with a procedure: the request is examined by a committee, the person or intermediary who placed the information is, where identifiable, given an opportunity to be heard, and a review committee examines the direction. A person aggrieved retains the remedy of a writ petition under Article 226. The Court also declined the invitation to condemn the confidentiality that the Rules attach to blocking requests, treating the subject matter, which includes defence, security and foreign relations, as a field in which the safeguard structure the Rules supply was sufficient. The discipline of the analysis should be noticed, because critics of this part of the judgment sometimes miss it: the Court did not hold blocking to be benign; it held that this blocking power, confined to constitutional grounds, attended by recorded reasons and a committee procedure, and open to challenge, is a different juridical object from a speech offence defined by undefined words, and the analysis of the Information Technology Act as a whole shows how much of the Act’s operative content sits in these ancillary provisions rather than in its offences.

Section 79 was saved by construction, and the construction is the part of the judgment with the widest daily operation, because it governs every platform takedown dispute in India. The problem was Section 79(3)(b) read with Rule 3(4). If an intermediary lost its immunity whenever an affected person complained in writing that hosted material was unlawful, then the intermediary became the judge of the complaint. The Court reasoned that an intermediary receiving millions of requests cannot be required to decide which are legitimate and which are not, applying legal standards of defamation, obscenity and the rest across the entire output of its users; and the predictable result of imposing that judgment on a private company, on pain of losing immunity for its users’ speech, was removal of lawful material whenever a complaint was made, because removal is always the safer course for the intermediary. The provision was therefore read down. Actual knowledge in Section 79(3)(b) means knowledge from a court order, which must direct the intermediary to expeditiously remove or disable access to identified material, or notification by the appropriate government or its agency, under the Article 19(2) framework, that unlawful acts relatable to Article 19(2) are going to be committed; and the notification or order must strictly conform to the subject matters in Article 19(2), since unlawful acts beyond those subject matters cannot form any part of the section. Rule 3(4) was read down in the same manner, so that the knowledge which obliges an intermediary to act within thirty six hours is knowledge through a court order and not knowledge supplied by a private complainant. With those two caveats, the Court refrained from striking down Section 79(3)(b). The Court noted, as comparative support for the construction, that the view it adopted had gained acceptance in other countries, with Argentina in the forefront: the adjudication of unlawfulness by a court or an accountable public authority, rather than by the host on private notice, was treated not as an Indian eccentricity but as the emerging answer of legal systems confronting the same volume problem. The reading preserves the immunity scheme by relocating the judgment of unlawfulness from the platform to the two institutions the Constitution trusts with it, a court and a government acting within Article 19(2), and it is the foundation on which every later argument about platform liability in India has been built.

What the reading displaced should also be understood, because the pre judgment operation of Rule 3(4) was a working system of private censorship with legal compulsion behind it, and its ordinariness is what made it dangerous. Any affected person could write to an intermediary asserting that hosted material fell within the prohibited categories of Rule 3(2), a list that ran from the blasphemous and defamatory to the merely harmful and disparaging, and the intermediary then had thirty six hours to work with the user and disable the material, on pain of the consequences that attended a failure of due diligence under the immunity scheme. The intermediary received no evidence standard, no hearing for the material’s author in its defence, no appeal, and no indemnity for a wrongful refusal to remove. Its incentives pointed one way. Removing the material cost it a user’s goodwill; refusing cost it the immunity on which its business model rested, in a legal regime whose other speech offence punished annoyance. The predictable result, documented in the arguments before the Court, was removal on complaint as a default business practice, and the creation of a takedown jurisdiction exercised by unaccountable private actors at the instance of any complainant, including complainants whose objection was to criticism of themselves. It is against that baseline that the modesty of the Court’s intervention should be measured. The judgment did not give platforms a charter and did not immunise them against the law; it held only that the trigger for compelled removal must be an adjudication or an accountable public act, and in doing so it converted an invisible private process into the first question of Indian internet law, which is always, after 2015, the question of who decided that the material was unlawful.

The blocking machinery whose existence carried Section 69A to safety deserves to be understood as a working system and not as a list of safeguards, because the difference between the two provisions in operation is the difference the judgment rewards. A blocking request is made in a prescribed form and placed before a committee, and the officer who chairs the process is the Designated Officer, a senior officer of the ministry administering the Act. The committee includes representatives drawn from the ministries and agencies the Rules specify, and its task is to examine the request against the grounds in the section. Where the person who created the information, or the intermediary hosting it, can be identified, the Rules contemplate that they be heard. There is an emergency route for cases that cannot await the committee, whose product must still return to the committee and to review. Above the whole process sits a review committee, constituted under the telegraph rules framework, charged with examining blocking directions at intervals. Every stage generates a record, and the reasons for a direction must be recorded in writing, which is what makes the direction amenable to a writ challenge in which a court can ask the questions the committee asked and test its answers. Whether this machinery is sufficient is the subject of the second criticism set out below; that it exists, is staffed, is documented and is examinable is the fact pattern on which the judgment’s third holding rests, and it has no counterpart anywhere in the design of Section 66A, under which the operative decision was taken by an officer registering a first information report, with no committee, no reasons addressed to the speaker, and no review other than the criminal trial itself.

Section 118(d) of the Kerala Police Act fell with Section 66A and for the same reasons in shorter compass. A provision that punishes causing annoyance to any person through a communication device is a speech offence whose operative word is a reaction, undefined and unbounded, and it could no more be fitted into the heads of Article 19(2) than the central provision could. Its inclusion in the judgment matters beyond Kerala, because it establishes that the analysis is not a peculiarity of the internet statute or of the Union legislature: a state legislature enacting the same vice in a police Act produces the same invalidity, and the discipline of Article 19(2) binds every legislature and every medium alike.

Two instruments of defence remain to be accounted for, because the judgment’s treatment of them governs every later case in which a government defends speech legislation by pointing away from the text. The first is the presumption of constitutionality, the working rule that a court approaches a statute expecting validity and resolves doubts in the legislature’s favour. The Court did not displace the presumption; it specified its limits in a way that deprives it of most of its force in a case of this kind. The presumption assists a provision whose meaning is doubtful between a constitutional and an unconstitutional construction, and it reflects a proper respect for a coordinate legislature’s assessment of the needs its law addresses. It does not assist a provision whose coverage is plain and whose defect is the coverage itself, since respect for the legislature cannot extend to attributing to it an intention to enact a different and narrower law than the one its words produce, and it carries no weight at all on the question the presumption is usually deployed to answer, which is whether the legislature’s purpose was legitimate: the Court accepted throughout that the harms the Union invoked, abuse, harassment and disorder online, were real and were legitimate objects of legislation. What failed was not the purpose but the instrument, and on instruments the Constitution’s demands are not a matter of deference but of text.

The second instrument is the assurance, and its rejection is the part of the judgment whose wisdom the afterlife has most completely vindicated. The Union invited the Court to sustain the provision on the faith of restrained administration, supported by the advisory that filtered arrests through senior officers. The Court declined the invitation on principle: the validity of a law falls to be judged by what it authorises, and a power to punish protected speech is not made constitutional by the present holder’s promise to use it sparingly, because the promise binds nobody, survives nobody, and is invisible to the citizen who must decide, at the moment of speaking, whether the text makes her a criminal. The advisory itself, on this reasoning, was evidence against the provision rather than for it, an official acknowledgment that the unfiltered text could not be trusted in the hands the law delivers it to. The reader of the implementation record will notice the exactness of the trap the judgment thereby avoided and the state walked into regardless. Had the provision been upheld on assurances in 2015, its continued use after the hypothetical judgment would have been a breach of faith by the administration. Struck down, its continued use was a breach of the Constitution by the same administration, undeterred by the clearest declaration the Court can issue. The variable between the two outcomes is not the solemnity of the pronouncement but the physical and institutional path between the pronouncement and the police station, and the judgment’s refusal to trade in assurances reads, in retrospect, as the Court declining to pretend that the path was shorter than it is.

The consequence of the first holding requires separate statement, because the article’s second half depends on it. A post constitutional law that violates a fundamental right is, under Article 13(2) of the Constitution, void to the extent of the contravention, and the established understanding of the Court’s declaration is that such a provision is void ab initio: it is treated as though it had never been enacted. The practical meaning for Section 66A was categorical. No fresh case could be registered under it, because there was no offence to register. No pending prosecution under it could continue, because there was no law under which the accused stood charged. No conviction under it could be sustained. The doctrine through which declarations of this kind acquire their force, and the limits of judicial review of legislation more generally, are treated in the series article on the basic structure doctrine, which owns that ground; what matters here is the exactness of the legal position the judgment created on 24 March 2015. On one side of that date there was, in the eye of the law, nothing: no provision, no offence, no power. On the other side there was a fully printed section of a working statute, reproduced in every copy of the bare Act, cited in first information reports as a matter of routine station house drafting, and framed in charge sheets before trial courts that took their texts from the same printed page. The judgment changed the law completely and the page not at all. Which of those two facts would govern the conduct of the institutions that used the provision was a question the law did not ask, and it turned out to be the only question that mattered.

The Judgment in the Line of the Speech Cases

No constitutional judgment is written on a blank page, and this one’s authority is best measured by the line of cases it organises. The Court’s speech jurisprudence begins with the two decisions of 1950, Romesh Thappar v. State of Madras and Brij Bhushan v. State of Delhi, in which bans imposed in the interest of public order as then unqualifiedly understood were struck down, provoking the First Amendment, which inserted public order into Article 19(2) together with the requirement that restrictions be reasonable and the head of incitement to an offence. The constitutional settlement that emerged is the one Shreya Singhal enforces: the heads are generous in subject but strict in fit, and a restriction’s claim to a head is examinable by the court. Within that settlement, the sedition case, Kedar Nath Singh, fixed the technique of saving construction; the Lohia case fixed the proximity requirement; the Rangarajan case fixed its metaphor and applied the settlement to a work of art, holding that criticism of government and its measures, however strongly worded, is not a ground for suppression in a democracy that takes its commitments seriously; and the defamation and privacy cases of the following decades, among them R. Rajagopal v. State of Tamil Nadu, developed the idea that a legal regime can violate speech without prohibiting it, by making its exercise too hazardous to undertake, the Indian root of the chilling effect that Shreya Singhal would name and operationalise.

Read against that line, the judgment of 2015 is a synthesis with two genuine additions. The first addition is the explicit ordering of discussion, advocacy and incitement as the spectrum along which every speech case is to be located before the Article 19(2) analysis begins. The materials of the ordering are old, in Kedar Nath Singh and the American authorities the Court cited, but the ordering itself, stated as doctrine and applied mechanically to a modern statute, is the judgment’s contribution, and it is the reason the case is taught as the organising authority for the subject rather than as an episode about a bad section. The second addition is the elevation of vagueness from an occasional argument to a principal ground of decision in a speech case, with a full comparative apparatus. Earlier Indian decisions had noticed vagueness, in the law of goondas, in preventive detention and in the Kartar Singh treatment of the anti terrorism statute, but the doctrine had not previously carried the destruction of a major speech provision. After 2015 it is the argument every challenger leads with and every drafter must design against.

The comparative dimension of the judgment should be characterised accurately, because it is often misdescribed as an importation of American free speech law wholesale. The Court used the American authorities for the doctrines they had developed in their most considered form, vagueness, overbreadth and the chilling effect, and for the vocabulary of incitement that the American cases from Schenck onward had refined; it did not adopt the American premise that speech restrictions are presumptively suspect in the abstract, since the Indian text itself enumerates the permissible restrictions and the whole of the Court’s method is to work from the enumeration. The judgment is thus comparative in its tools and textual in its foundations, a combination that explains both its influence and its durability: later benches that disagree with American doctrine can still apply Shreya Singhal, because its operative propositions are propositions about Article 19(2), and the comparative material is scaffolding rather than structure.

It is equally worth recording what the decision did to the burden of justification in speech litigation. Before it, a defending government could substantially rely on the presumption of constitutionality, on the breadth of the public order head, and on the doctrine that abuse is no ground of invalidity, and the challenger carried the practical burden of displacing all three. The judgment does not formally reverse any presumption, but it reorders the argument so completely that the practical burden moves. Once the challenger has shown that a provision reaches discussion and advocacy, that its terms are undefined, and that its documented operation has caught protected speech, the government must answer with a head, a proximate connection and a definable standard, and assurances about administration are ruled out of the answer in advance. Governments drafting or defending speech legislation since 2015 have had to write to that standard, and the cases in which later provisions have survived challenge are, without exception, cases in which the provision could produce the head, the proximity and the definition that Section 66A could not.

The Doctrine as the Judgment Leaves It

The doctrinal residue of the case can be organised under four heads. The first is the location of the boundary of permissible restriction. Before 2015, Indian speech law contained the materials of the discussion, advocacy and incitement distinction, in the sedition cases and in the public order cases, but it had not stated the distinction as the organising principle of Article 19 analysis nor applied it to destroy a provision on the ground that the provision criminalised the discussion side of the line. Shreya Singhal did both. A speech offence in India must after this judgment be drafted to the incitement end of the spectrum and tied to a head of Article 19(2) with a proximate connection between the words and the harm, or it must expect to meet the argument that ended Section 66A. The judgment accordingly became the first authority cited in every later challenge to a speech offence, and its framework is the one against which proposals for new offences concerning online content are publicly measured.

The second head is the method of the challenge itself. The judgment consolidated vagueness and overbreadth, doctrines with American roots, as working instruments of Indian constitutional adjudication, in tandem with the older indigenous instruments of Article 19(2) fit and proximate nexus. A challenger after 2015 pleads in a standard order: the provision does not answer to a head of Article 19(2); if it does, its connection to the harm is remote; its terms are so indefinite that a citizen cannot regulate conduct by them and an official can enforce them arbitrarily; and it catches protected speech in such proportion that its ordinary operation chills the exercise of the right. That pleading structure, and the burden it places on a defending government to show a defined offence, a specified ground and a proportionate reach, is the judgment’s practical bequest to litigation. The relationship of this structure to the proportionality method that the Court would elaborate for rights adjudication more generally, in the privacy decision and the Aadhaar litigation that followed, is traced in the article on the Puttaswamy privacy and Aadhaar judgment; the two decisions are usually read together, Shreya Singhal supplying the speech specific grammar of grounds, proximity and definiteness, and Puttaswamy supplying the general grammar of legitimate aim, suitability, necessity and balance.

The third head is the treatment of intermediaries, which after 2015 became the operative takedown law of India. The reading down of Section 79(3)(b) and Rule 3(4) meant that a private complaint, however numerous or however organised, could not by itself oblige a platform to remove material on pain of losing immunity, and that removal under legal compulsion required a court order or a government notification within Article 19(2). Every subsequent dispute about platform responsibility, and every later set of intermediary rules, has had to position itself against that holding, either by working within it or by constructing a machinery that supplies the adjudicatory step the judgment required. In practice the holding produced a two channel system of removal that persists as the background law of the Indian internet. Material comes down under legal compulsion through the courts, on orders that identify it and can be contested by the person who posted it, or through the blocking machinery, on the government’s direction within the constitutional grounds. Material also comes down, in far greater volume, under the platforms’ own terms of service, which the judgment did not touch and which operate as private law; and the distinction between the two, between removal the Constitution has authorised through an adjudicatory route and removal a contract permits without one, is the distinction the judgment’s critics and defenders both treat as the live question in the field. A reader who understands that distinction, and understands that the judgment governs the first channel and not the second, understands both the reach of Shreya Singhal and its boundary, which summary accounts of the case, in both celebration and criticism, habitually blur. For the examination candidate and the teacher, the case is among the most examined in the constitutional law syllabus, and a reader assembling the three holdings into teaching material can lay the file out and check the finished deck in an online PowerPoint viewer before the class, since the distinction the judgment draws is one that survives or dies on the precision of the slide that states it.

The fourth head is negative, and it is the one the litigants of 2015 would have found most surprising. The judgment decided nothing about the constitutionality of speech offences drafted with the definiteness it required. It did not hold that online harassment, targeted abuse, threats or incitement cannot be criminalised. It held that these things cannot be criminalised by words that also criminalise everything else, and it left the legislature free, as legislatures remain free, to attempt the drafting the provision had not attempted. The space the judgment leaves open is the space in which the debate about a successor provision has been conducted ever since, and the debate’s difficulty is the judgment’s doing: any successor must name its harm, define its terms, station itself under a head of Article 19(2), and show a proximate connection between the words punished and the harm named, and a legislature that does all of that discovers that the resulting provision looks very different from the one that was struck down.

The drafting standard the judgment sets can be stated as a positive specification, and stating it positively is the most useful service an account of the case can perform for the reader who drafts, advises on or campaigns about legislation in this field. A speech offence that expects to survive must first select its ground from the eight heads of Article 19(2) and be able to say, of its whole coverage and not of a chosen example, that the coverage falls within the ground. It must define its operative terms by elements that can be proved, a specified act, a specified mental element and, where the ground requires it, a specified tendency or consequence, and it must avoid standards that consist of the reactions of an audience. It must build the proximity the public order cases require into its text, by requiring the connection between the words and the disorder, danger or offence, rather than leaving the connection to be assumed. It must grade its response, distinguishing the casual utterance from the campaign and the communication between persons from the address to the public, so that its punishment does not fall with equal weight on the whole of its coverage. And it must locate the judgment of lawfulness in institutions designed to make it, since a provision that can be set in motion by any complainant and applied by any officer, with the first legal assessment occurring in a courtroom after arrest, will reproduce, whatever its verbal care, the practical vice that destroyed Section 66A. Measured against that specification, the recurring proposals for successor provisions can be assessed without rhetoric, and their difficulties are visible in advance: the harms that prompt them, stalking, targeted harassment, coordinated abuse, are real and definable, but they are definable precisely by elements, of repetition, of targeting, of threat, that the framers of Section 66A declined to write, preferring a vocabulary of reaction that the Constitution has since been held not to accept.

The judgment’s place in later litigation confirms the specification from the other side. When the Kerala High Court held in Faheema Shirin v. State of Kerala that access to the internet partakes of the freedoms the Constitution protects, it reasoned within the world Shreya Singhal had made, in which the medium is constitutionally salient rather than constitutionally marginal. When the Supreme Court considered the restrictions imposed in Kashmir in Anuradha Bhasin v. Union of India, the case was argued and decided in the grammar of Article 19(1)(a) applied to expression through the internet, with the state required to justify restrictions by their grounds and their necessity. And in the continuing stream of challenges to speech offences in the penal and technology statutes, the pleadings follow the order the 2015 judgment standardised: fit with a head, proximity to the harm, definiteness of the terms, proportion of protected speech caught. A judgment’s doctrinal success is measurable by whether the arguments in the next generation of cases are unintelligible without it, and by that measure the case succeeded completely within five years of its delivery, which makes the failure recorded in the second half of this article the more striking: the profession absorbed the judgment at once, and the police station did not absorb it at all.

How did the judgment treat the argument that the internet is a separate medium?

The Court accepted that the legislature could treat the internet as a distinct field of regulation, given the reach, speed, anonymity and permanence of the medium, and it rejected the argument that legislating for online speech alone was, without more, a forbidden classification.

But a separate field is not a separate Constitution: the medium’s characteristics could not enlarge the grounds of restriction, and its dangers could not convert annoyance into a head of Article 19(2).

The judgment also left the law of sedition exactly where Kedar Nath Singh v. State of Bihar had left it in 1962, and it used that case as its model of a speech provision saved by construction. In Kedar Nath Singh the Court upheld Section 124A of the Penal Code by confining it to acts involving incitement to violence or an intention or tendency to create public disorder. Shreya Singhal treated that reading as settled law and reasoned from it: incitement is the point at which Article 19(2) becomes available, and a provision confined to incitement can stand where a provision punishing discussion and advocacy cannot.

Sedition is, in fact, the comparison that does most to display the judgment’s logic. The sedition provision, for all the criticism it attracts, names a target, the bringing of the government into hatred or contempt or the exciting of disaffection, and a century of judicial construction had supplied it with a limiting gloss that a court could adopt as the provision’s settled meaning. Section 66A named only reactions, and no settled construction existed or could be assembled, because there was no stable meaning to settle on. The judgment’s refusal to perform for Section 66A the rescue performed for Section 124A is therefore not an inconsistency but a statement about the preconditions of rescue: a court can confine a provision to a meaning its words and its history will bear, and it cannot manufacture a provision out of words that bear every meaning and therefore none.

What the judgment did not do is almost as important for its later career. It did not hold that the internet is constitutionally irrelevant to speech regulation; it accepted the medium could be separately classified and separately legislated for, provided the legislation observed the same restrictions discipline. It did not constitutionalise a right to anonymity, a right to be forgotten, or any of the later digital rights claims sometimes attributed to it in summary writing. It did not disturb the defamation, contempt or obscenity laws, whose fit with Article 19(2) it treated as supplied by the heads themselves. And it did not find in Article 19(1)(a) a general immunity for false speech: the provision it destroyed punished falsehood only as one route into annoyance, and the Court’s objection was to the destination, not to the law’s capacity, under other instruments, to reach fraud and deception. The judgment is best understood as a judgment about drafting discipline under a closed list of grounds, and its authority is at its strongest when it is cited for exactly that.

The Two Criticisms, Stated in Full

The judgment attracts criticism from two directions, and the register of this series requires both to be stated at full strength and neither to be adjudicated here. The first criticism comes from the direction of the harm the provision was aimed at. On this view, Section 66A, for all its faults, was the only provision addressed specifically to abuse and harassment delivered through a communication device, and its destruction left a gap. The general criminal law, it is said, fits the harm awkwardly: the Penal Code’s provisions on insult intended to provoke a breach of the peace, on criminal intimidation, on defamation and on words intended to outrage religious feelings were drafted for other settings, carry their own thresholds and sanction requirements, and do not map onto the experience of a person subjected to sustained online abuse, much of it anonymous, much of it falling short of any single offence while amounting in aggregate to a campaign. The further criticism is institutional: that the Court, which had read down other provisions in the past and which read down Section 79 in the very same judgment, could have narrowed Section 66A’s expressions by construction, supplying the thresholds the provision lacked, rather than removing the instrument altogether. The judgment’s own answer to this criticism is on its face and is part of the record: that the vice was vagueness and overbreadth of a kind that construction cannot cure, because there was no valid core to construe the provision down to, and a court performing the narrowing would have been enacting the offence it pretended to interpret. A reader deciding between the criticism and the answer should notice what each assumes about the function of a court, the criticism assigning the court a curative role over defective speech legislation, the judgment assigning it a boundary keeping role and leaving the cure to the legislature. The institutional history since 2015 gives the criticism a concrete exhibit, and fairness requires it to be recorded. Within government, the response to the gap was not a drafted Bill laid before Parliament but a committee: the committee chaired by T. K. Vishwanathan, a former law secretary, which examined the framework for a fresh provision in the territory Section 66A had vacated. When its recommendations surfaced, through the efforts of the Internet Freedom Foundation rather than through official publication, they proposed new offences in the same territory, and the Foundation’s criticism of the process, that the committee had taken no public input, had no civil society membership and had not published its recommendations, became part of the debate about the gap itself. The episode thus cuts in both directions, and this article leaves it uncut: it shows the state moving, as the first criticism would wish, to fill the space the judgment cleared, and it shows the filling attempted, as the judgment’s method would predict to be fatal, by a route that avoided the publicity and the definitional discipline on which a defensible provision would depend. No successor provision has been enacted, and the gap the criticism identifies remains, years later, a gap in fact as well as in law.

The second criticism comes from the opposite direction and attacks the part of the judgment that upheld power rather than the part that struck it down. On this view, the vigilance the Court applied to a speech offence was not applied to the blocking power, and the contrast is difficult to defend. Blocking under Section 69A is done by the executive, through a committee of officials; the orders are, under the Blocking Rules, kept confidential; the person whose material is blocked is entitled to a hearing only where identifiable and in the manner the Rules allow; and the proceeding has none of the publicity that disciplines a criminal prosecution, in which the charge must be stated, the evidence led and the reasoning exposed to appeal. A regime in which the state can cause speech to disappear without a public order, a stated case or an affected speaker in the room is, on this criticism, at least as dangerous to the freedom as a vague offence that must at least be defended in open court, and the safeguards the judgment relied on, recorded reasons and committee review, are safeguards internal to the very executive whose action they are meant to check. The judgment’s answer, again on its face, is that the power is textually confined to the Article 19(2) grounds, that the Rules supply procedure, and that the writ remedy lies; the criticism replies that confinement to grounds is exactly what Section 66A’s defenders claimed for the heads the provision gestured towards, and that the difference in outcome rests on trusting the procedure of the blocking regime more than the procedure of the criminal courts. This article adjudicates neither criticism. It records that both are made by serious commentators on the judgment, that each identifies a real feature of the case, and that the two criticisms cannot both be answered by the same judicial technique, since the first asks the Court to do more construction and the second asks it to trust construction less.

On the afterlife, a different discipline of statement applies, and it is imposed by the record itself. The documented explanation of the continued prosecutions, given in the research and repeated by the Attorney General in court, is mechanical: the provision remained in the printed statute, and institutions that work from the printed statute used it. Whether any particular registration was deliberate defiance, local opportunism, or the routine copying of a section number from the book on the station house desk is not something the record establishes case by case, and this article attributes no motive beyond the mechanism the record supports, which is in some ways the more disturbing finding, since a mechanism requires no villain and is correspondingly harder to switch off.

Struck Down and Still Used: The Afterlife of Section 66A

The afterlife of the provision began in the manner of a rumour that turned out to be a dataset. Reports accumulated of first information reports citing Section 66A years after March 2015, of arrests, of charge sheets and of trials proceeding in district courts on a charge that did not exist. In October 2018, Apar Gupta and Abhinav Sekhri published a study, Section 66A and Other Legal Zombies, which gave the phenomenon a name and a mechanism. Their account, on which much of the later litigation was built, identified a signal failure between the court that declares a law unconstitutional and the institutions that must act on the declaration. The judgment had not been carried into the text of the statute: Parliament had enacted no amendment removing the provision, commercial publishers of the bare Act printed it with a footnote recording that it had been struck down, and the official text carried no equivalent signal to the officer or the magistrate who consulted it. Dissemination of the judgment through official channels had been limited, and no institution owned the task of auditing pending cases for the dead charge. A provision can therefore be, in their term, a zombie: dead in constitutional law and ambulant in the criminal process, because the criminal process runs on printed texts, station level practice and file level inertia rather than on the law reports.

The People’s Union for Civil Liberties, which had been a petitioner in the original batch, returned to the Court. In an application filed in January 2019 it recorded that more than twenty persons had been prosecuted under the provision after it was struck down, and sought directions to make the judgment effective: that copies of the judgment be circulated to the governments and courts that would have to apply it, and that pending cases be identified and disposed of. At the hearing in January 2019, a bench of Nariman J. and Vineet Saran J. was reported as warning in the strongest terms that officials responsible for arrests under the provision would themselves face imprisonment, and the Union was directed to respond. On 15 February 2019 the Court disposed of the application with the directions the petitioners had principally sought. Copies of the Shreya Singhal judgment were to be made available by every High Court to all the district courts under it. The Centre was to make copies available to the Chief Secretaries of all States and Union territories, and the Chief Secretaries were in turn to sensitise the police departments of their States, sending the judgment onward to each Director General of Police. The design of the order is worth pausing on. It is an order about the transmission of information through the administrative system, addressed to the exact failure the research had identified: the Court treated its own judgment as a message that had not arrived, and used its authority to build the channel through which the message was to travel.

The Union had, in January 2019 and under the pressure of the same application, acted in parallel: an advisory went to the State governments asking them to furnish the data of cases under the provision, and the responses illustrated the information problem from the other side, since most States did not respond at all and those that did, Kerala among them with a return of eighteen cases, disclosed only fragments of a picture no government had assembled. The episode also produced judicial echoes below the Supreme Court. The Allahabad High Court was reported, in terms that travelled through the legal press, as expressing amazement that first information reports were being registered under Section 66A despite the provision having been declared ultra vires, a reaction that captures the position of the district judiciary’s supervisory tier: the High Courts, which had received the judgment in the ordinary way and applied it in their own courts, were discovering from litigation before them that the courts and stations below were working from a different book. The significance of the 2019 round is that it converted a suspicion into an institutional admission. After February 2019, no participant in the system could plausibly describe the continued prosecutions as unproved; the Supreme Court had directed the judgment’s circulation on the express footing that prosecutions were continuing, and the question that remained was only whether circulation would work.

The 2019 order is also, in the taxonomy of Indian public law, a document of some novelty, and its novelty is diagnostic. Implementation directions of this kind, in which a constitutional court moves from declaring rights to engineering the administrative transmission of its own decision, have their closest relatives in the continuing mandamus tradition, in which the Court retains a matter and supervises compliance over years. What distinguishes the Section 66A proceedings is that the supervision was required not against a recalcitrant government defending a policy, but against entropy: no institution before the Court in 2019 or 2021 defended the use of the provision, the Union’s own law officer explained the failure as a reading problem, and the State governments’ responses, when compelled, conceded the facts. The case therefore isolates a variable that constitutional litigation usually confounds with opposition. It shows what compliance looks like when nobody disagrees, which is that compliance still does not occur by itself, because a judgment is an event in the legal system and not an instruction in the administrative one, and the two systems are connected by channels, the amendment of the statute book, the revision of manuals and training, the audit of live files, that somebody must be directed, or must volunteer, to operate. In this case every one of those channels had to be litigated into existence separately.

Cases continued. The measurement of the failure was undertaken by the Internet Freedom Foundation, an internet rights organisation, working with CivicDataLab, a data laboratory, in a project named the Zombie Tracker, launched on 15 January 2021. The Tracker was built from district court records on the eCourts platform across eleven States, with case records collected from the commencement of the provision in October 2009, and it converted anecdote into arithmetic. Its figures, as placed before the Supreme Court in the proceedings that followed, were these. In the eleven States covered, 229 cases under Section 66A were pending when the provision was struck down in March 2015. After the provision ceased to exist, the police in those States lodged 1,307 fresh cases under it. As on 10 March 2021, 745 of those cases were still pending and active before the district courts, with Maharashtra accounting for 381 of them and Uttar Pradesh for 245. The comparison in the brief of every account of this episode is exact and it is not an artefact of counting: more prosecutions were instituted under Section 66A in the six years after the Supreme Court destroyed it than had been instituted in the period of its validity that the same records capture, and the rate of institution rose, in the description later used in the proceedings, fivefold. The researchers were careful, and this article is careful, about the limits of the exercise: the eCourts records are incomplete and unevenly digitised, the coverage was eleven States and not the Union, and the true all India figure for post judgment prosecutions is therefore larger than the counted one, by an amount the method cannot state.

Two features of the counted record deserve emphasis, because they exclude the charitable explanations of the practice. The first is that the dead charge did not survive only as an appendage to live ones. The application placed before the Court recorded, as a distinct and disturbing finding, cases in which charges had been framed solely under Section 66A, so that the entire proceeding, the arrest, the remand, the charge sheet and the trial, rested on a provision that did not exist. A copying error in a multi section first information report might be corrected at the first hearing; a prosecution framed solely on the dead section is a proceeding that passed through the hands of the registering officer, the investigating officer, the prosecutor and the court that took cognisance, at none of which points did the system interrupt it. The second feature is the identity of the accused. The Tracker’s launch was marked by the account of Zakir Ali Tyagi, a young man from Uttar Pradesh who had been charged under the provision after the judgment for a social media post, and whose case the researchers used to illustrate what the aggregate figures mean at the level of a life: an arrest, a period in custody, and a criminal proceeding for an offence that the Constitution had already abolished, defended at the cost and over the years that criminal proceedings in the district courts consume. The researchers’ methodological modesty about their totals is compatible with, and indeed sharpens, the conclusion the Court drew from them: the counted cases are a floor and not a ceiling, and every case in the count is a completed constitutional violation, not a statistic with a margin of error in its moral character.

The two implementation rounds, of 2019 and of 2021 to 2022, display an escalation logic that is itself instructive about how declarations fail. The first round assumed the failure was one of knowledge and prescribed information: circulate the judgment. The second round abandoned that assumption, because the information had been circulated and the practice had not changed, and prescribed verification and excision instead: count the cases, name the jurisdictions, delete the charge from the files, instruct the force, and annotate the book. Between the two rounds lies the lesson the researchers had been pressing from the start, which is that an unconstitutional provision embedded in the working texts of the criminal process is not dislodged by being known to be unconstitutional at the level of the system that reads judgments. It is dislodged at the level of the file, the register and the printed page, and dislodging it requires an institution willing to work at that level, with data. That the data had to be built by a civil society organisation, from a public judicial database, because no organ of the state counted the prosecutions it was conducting under a dead law, is the single most damning administrative fact in the record, and it was stated in those terms by nobody more plainly than by the structure of the proceedings themselves, in which the Court’s information about its own legal system arrived by annexure to a non governmental application.

Armed with that record, the People’s Union for Civil Liberties returned to the Court a second time in 2021, with an application that made a series of recommendations for ensuring implementation: that the Centre collect data from every State on first information reports, investigations and pending cases under the provision; that the Registry communicate the judgment to all district courts through the High Courts so that cognisance be taken of it in any proceeding in which the provision was invoked; that High Courts collect information on pendency from their district courts; that the Centre advise every police station against registering cases under the provision; and that the public be informed, including through newspapers, that Section 66A had been struck down and was no longer law. On 5 July 2021 the application was heard by a bench of Nariman J., K. M. Joseph J. and B. R. Gavai J. Nariman J., the author of the 2015 judgment, is reported to have opened with the words that gave the hearing its public character: amazing, the Court said; what is going on is terrible. Notice issued to the Union. It was then that the mechanism of the whole affair was stated in open court by the government’s own law officer. Attorney General K. K. Venugopal, appearing for the Union, explained that Section 66A is still in the Act; that when the Act is perused, the section is there, and in the footnote it is written that the section has been struck down; and that when a police officer has to register a case, he sees the section and registers the case without reading the footnote. His suggested repair was typographical: place a bracket against the section recording that it has been struck down, and put the entire extract of the judgment in the footnote. The bench’s response is the sentence the brief of this article takes as its centre, and it is reported in substance rather than ornament: a police officer would not read the footnote. The exchange deserves its prominence in the literature of this case because it identifies, with the agreement of both sides of the courtroom, the physical location of a constitutional failure, which is the page of a printed book and the reading habits of the person holding it.

The Union’s counter affidavit, filed in the weeks that followed, placed the operational burden on the States and Union territories, observing that law enforcement agencies act under the provisions of law and that it was equally their responsibility to comply with the judgment. The Ministry of Home Affairs, on 14 July 2021, wrote to the State governments and police chiefs directing that no fresh cases be registered under Section 66A and that cases already registered be withdrawn. On 2 August 2021 the Supreme Court sought responses from all States, Union territories and High Courts. The replies, when they came in, did not close the matter. In September 2022 the Court expressed its displeasure on being told, through affidavits filed on behalf of eight High Courts and six States, that 470 cases under the provision were still pending within the responding jurisdictions alone. On 12 October 2022, a bench of Lalit C.J. and Ajay Rastogi and S. Ravindra Bhat JJ. disposed of the application with the directions the episode had been travelling towards for three and a half years. No citizen is to be prosecuted under Section 66A, the Court directed, and in every case in which a citizen faces prosecution for an alleged offence under it, the reference to and reliance on the provision shall stand deleted, with other offences alleged in the same case unaffected. Directors General of Police, Home Secretaries and the competent officers of the States and Union territories were directed to instruct the entire police force not to register any complaint under the provision. And in a direction that completes the logic of the Attorney General’s own explanation, the Court directed that whenever any publication, government, semi government or private, publishes the Information Technology Act and quotes Section 66A, readers must be adequately informed that the provision has been pronounced unconstitutional. Seven and a half years after the judgment, the legal system had moved from declaring the law, to circulating the declaration, to counting the disobedience, to ordering the deletion of the charge from every live file and the annotation of every printed copy.

The form of the 2022 directions repays attention, because it shows the Court at last addressing the proceeding and not merely the provision. A prosecution under a struck down section rarely travels alone; first information reports in these cases characteristically club the dead section with living offences under the Penal Code, insult intended to provoke a breach of the peace, intentional insult, statements conducing to public mischief, criminal intimidation, and the clubbing is what allowed so many files to proceed undisturbed, since the file as a whole appeared regular to every reader who did not isolate the dead charge within it. The direction that the reference to Section 66A stand deleted, with the other offences expressly preserved, is drafted for that reality: it neither quashes proceedings wholesale nor leaves the dead charge embedded, but excises it file by file, which is the only unit at which the failure ever existed. The companion direction, that the entire police force be instructed not to register fresh complaints, addresses the other end of the pipeline, and the publication direction addresses the source document itself. Read together, the three directions map exactly onto the three sites of the failure the research had identified, the station, the file and the book, and their completeness as a map is the measure of how much litigation was required to teach the system where its own failure lived.

One further actor in the afterlife should be named, because its existence is a condition of everything recorded above and a lesson in its own right: the eCourts database itself. The counting that converted anecdote into proof was possible only because the district judiciary’s case records had been digitised and published, imperfectly, unevenly and with fields so unstandardised that the research team recorded standardisation as its single largest labour, and wrote to the Supreme Court’s e-Committee in May 2021 setting out the difficulties the platform had presented and the improvements that would let others repeat the exercise. A legal system that publishes its working records, even badly, arms its own critics with the means of audit, and the contrast inside this story is exact: the provision survived because one text, the statute book, was read without its annotation, and it was exposed because another text, the court record, could be read in bulk by anyone who cared to count. Transparency of the second kind is what eventually corrected opacity of the first, and any account of the case that omits the database mistakes a story about institutional design for a story about individual fault.

The position of a person against whom the provision is invoked follows from this record and can be stated plainly. No offence under Section 66A exists, and none has existed since 24 March 2015. A first information report that cites it cites a provision the Supreme Court has struck down, and the October 2022 directions require the reference to stand deleted from any pending proceeding and forbid the registration of any fresh complaint. What a person in that position should do in a specific case, which forum to approach and in what sequence, depends on the facts and the stage the proceeding has reached; the general account given here is legal information about the state of the law and is not a substitute for advice on an individual case, and anyone facing a live proceeding under the provision should take that advice promptly, since procedural stages, once crossed, are not always recoverable. The provision itself, finally, has not been repealed. No amending Act of Parliament has removed Section 66A from the text of the Information Technology Act, 2000, and a reader who opens the statute will find it printed there, in sequence between Sections 66 and 67, carrying the annotation that the October 2022 order requires to be adequate. The repeal that would end the matter by removing the words has not been enacted, and proposals examined within government for successor provisions on online abuse, including the work of the committee chaired by T. K. Vishwanathan whose recommendations were prepared without public consultation and were not published when made, have not produced a replacement section either. The law on the page, years after the law in the reports was settled, remains a section that exists and does not exist at once.

Stage	What the legal position was	What was happening on the ground
27 October 2009: Section 66A commences	A valid offence, punishable with up to three years of imprisonment, in the widest terms	Arrests for online speech begin, including of persons who had criticised public figures
24 March 2015: Shreya Singhal delivered	Section 66A struck down in its entirety, void in law; pending prosecutions cannot continue	The provision remains printed in every copy of the Act, with at most a footnote recording the judgment
2015 to 2018: the quiet years	No offence under the provision exists anywhere in India	Fresh first information reports and trials continue; the Gupta and Sekhri study of 2018 documents the practice and names it a legal zombie
15 February 2019: first implementation order	The judgment is directed to be circulated to all district courts, Chief Secretaries and Directors General of Police	Registrations continue; the circulation order does not reach the station house as a change of practice
15 January 2021: the Zombie Tracker	The legal position is unchanged since 2015	District court records in eleven States show 1,307 fresh cases lodged after the judgment, against 229 pending when it was delivered; 745 remain pending in March 2021
5 July 2021: the second PUCL hearing	The bench records that the state of affairs is terrible and issues notice to the Union	The Attorney General explains the mechanism in court: the officer reads the section, not the footnote
14 July 2021: Home Ministry communication	States and police chiefs are directed to stop registering cases and to withdraw pending ones	Compliance is partial; affidavits in 2022 still disclose 470 pending cases in the responding jurisdictions alone
12 October 2022: final directions	No citizen is to be prosecuted; the reference stands deleted from every pending case; publications quoting the provision must state that it is unconstitutional	The provision remains in the Act, unrepealed, its practical life governed by directions about annotation rather than by removal of the text

A Footnote Is Not a Repeal

The namable claim of this article can be stated in its full form. A footnote is not a repeal. A provision declared void from its inception by the Supreme Court in 2015 continued to be printed in the statute book with a footnote recording that it had been struck down; police and subordinate courts, who work from the printed page and not from the law reports, continued to use it; and the measured result, on district court records across eleven States, is that more prosecutions were instituted under it in the six years after the judgment than were pending when it fell, at a rate the proceedings themselves described as fivefold. The practical life of a criminal provision in India depends, to a degree that constitutional lawyers trained on the reports do not naturally credit, on what appears in the bare Act. Everything in the Shreya Singhal story is consistent with that claim and nothing in it is explicable without it: not the first failure, which followed a judgment of exemplary clarity unaccompanied by any change to the text; not the shape of the remedy, which was at every stage a remedy about transmission, circulation, instruction and annotation rather than about doctrine; and not the Attorney General’s concession, which located the whole failure in the distance between a section and its footnote.

None of this diminishes the judgment, and the article’s verdict on the judgment itself should be stated as plainly as its verdict on the aftermath. As an exercise in constitutional adjudication, Shreya Singhal is complete: it identifies the governing principle, applies it provision by provision with a discrimination that produced four different results for four different texts, consolidates the doctrines of vagueness, overbreadth and chilling effect into the working law of speech, and writes the intermediary holding on which the takedown law of the country has since been built. Its limits are the limits of its instrument. A judgment speaks to the parties, to the courts below in their appellate capacity, and to the profession that reads it. It does not reprint the statute book, retrain the station house, or audit the district court file, and between 2015 and 2022 each of those tasks had to be demanded separately, by a civil liberties organisation that returned three times, by researchers who built a counting instrument out of public court records because no official count existed, and by benches that escalated from circulation to instruction to deletion as each lesser direction failed. The series to which this article belongs has traced statutes in force and not applied, and rules weaker than the Acts they serve. This case is the mirror image and the sharpest instance in the file: a provision not in force and applied anyway, whose mechanism of survival was that it was still printed, and whose history since 2015 is the standing refutation of the assumption that constitutional law is finished when the judgment is delivered. The judgment belongs to every Indian who writes, publishes, hosts or merely reads on the internet, whether or not its name is known to them, because the terms on which all of them do those things were reset on a single morning in March 2015 and have been contested on the ground the judgment marked out ever since.

Frequently Asked Questions

Q: What did the Shreya Singhal judgment decide?

The judgment, delivered on 24 March 2015 in Shreya Singhal v. Union of India, (2015) 5 SCC 1, decided four challenges. Section 66A of the Information Technology Act, 2000 was struck down in its entirety as violative of Article 19(1)(a) and not saved by Article 19(2). Section 69A and the Blocking Rules of 2009 were upheld as constitutionally valid. Section 79 was upheld with Section 79(3)(b) read down, and Rule 3(4) of the Intermediaries Guidelines Rules, 2011 read down in the same manner. Section 118(d) of the Kerala Police Act was struck down for the same reasons as Section 66A.

Q: On what ground was Section 66A held unconstitutional?

Section 66A was held to violate the freedom of speech and expression in Article 19(1)(a), and to fall outside every saving ground in Article 19(2). Its operative expressions, grossly offensive, menacing, annoyance, inconvenience, danger, obstruction, insult, injury, criminal intimidation, enmity, hatred and ill will, are undefined and open ended, correspond to no head of Article 19(2), and reach a very large amount of protected and innocent speech. The Court held the provision vague, because a person of ordinary intelligence could not know in advance what was prohibited, overbroad, because its coverage was structural rather than incidental, and productive of a chilling effect on speech the Constitution protects.

Q: What is the difference between advocacy and incitement in free speech law?

The Shreya Singhal judgment organised free speech doctrine around three distinct concepts. Discussion is the exchange of views on any subject. Advocacy is the promotion of a cause, however unpopular that cause may be. Both discussion and advocacy lie at the heart of Article 19(1)(a) and are protected even when the views expressed are offensive to the majority or to the government of the day. Incitement is different in kind: it is speech that urges the commission of an unlawful act in terms proximate to the act itself. Only at the level of incitement does Article 19(2) permit the state to intervene, and only on one of its enumerated grounds.

Q: What did the Court hold about the blocking power?

The Court upheld Section 69A of the Information Technology Act and the Blocking Rules of 2009 as constitutionally valid. The power to block public access to information was sustained because it is confined to the subject matters of Article 19(2) and adds nothing beyond them, because the government must record its reasons in writing, because the Rules route the request through a committee with an opportunity of hearing where the originator is identifiable, because a review mechanism exists, and because an aggrieved person may challenge a blocking direction in a writ petition under Article 226. The contrast with Section 66A is textual: a narrow power with recorded reasons and a procedure survived, where an undefined offence failed.

Q: How did the Court read down intermediary liability?

Section 79(3)(b) was read down so that the actual knowledge which costs an intermediary its immunity means knowledge from a court order directing removal of identified material, or notification by the appropriate government or its agency that unlawful acts relatable to Article 19(2) are going to be committed. Rule 3(4) of the Intermediaries Guidelines Rules, 2011 was read down in the same manner. A complaint by a private person therefore cannot by itself oblige a platform to remove material. The Court reasoned that intermediaries receiving millions of complaints cannot be made judges of which are lawful, and that requiring them to judge would lead to the removal of lawful speech whenever a complaint was filed.

Q: Why did prosecutions continue after Section 66A was struck down?

Because the provision stayed on the printed page. Parliament enacted no amendment removing Section 66A from the Information Technology Act, so the bare Act continued to carry it, with at most a footnote in commercial editions recording that it had been struck down, and no equivalent signal in the official text. Police officers and trial courts work from the printed statute, and the judgment’s circulation through official channels was limited, as the study by Apar Gupta and Abhinav Sekhri in 2018 and the Zombie Tracker data later documented. The Attorney General stated the mechanism in court in July 2021: the officer registering a case reads the section and does not read the footnote.

Q: What did the Supreme Court do about continued use of Section 66A?

The Court acted in stages. On 15 February 2019, on the application of the People’s Union for Civil Liberties, it directed that copies of the judgment be made available to all district courts through the High Courts and to all Chief Secretaries, who were to sensitise their police forces. In July 2021 it issued notice on a second application supported by the Zombie Tracker figures, and the Home Ministry then directed States to stop registering cases and withdraw pending ones. On 12 October 2022 it disposed of the matter with directions that no citizen be prosecuted under the provision, that the reference to it stand deleted from every pending case, and that police forces be instructed not to register fresh complaints.

Q: Has Section 66A been removed from the statute book?

No. Section 66A has never been repealed by Parliament and remains printed in the text of the Information Technology Act, 2000, between Sections 66 and 67. The Supreme Court’s judgment made it void in law from 24 March 2015, so it creates no offence and supports no prosecution, and the directions of 12 October 2022 require that publications quoting it inform readers that it has been pronounced unconstitutional. Removal of the words themselves would require an amending Act, and none has been enacted. The provision is the leading Indian example of a law that is dead in the reports and present in the book.

Q: When was the Shreya Singhal case decided and which judges decided it?

The case was decided on 24 March 2015 by a two judge bench of the Supreme Court of India. The bench consisted of Chelameswar J. and Nariman J., and the judgment was delivered by Nariman J. It disposed of a batch of ten writ petitions, the lead petition being Writ Petition (Criminal) No. 167 of 2012 filed by Shreya Singhal. Nariman J. also presided, six years later, over the bench that heard the People’s Union for Civil Liberties application about the continued use of the struck down provision, and recorded the Court’s dismay at what the record showed.

Q: What is the citation of the Shreya Singhal judgment?

The judgment is reported as Shreya Singhal v. Union of India, (2015) 5 SCC 1, with the parallel report AIR 2015 SC 1523. It was delivered on 24 March 2015 by a bench of Chelameswar J. and Nariman J., the judgment being authored by Nariman J. The operative conclusions are collected in the final paragraph of the judgment, which records the striking down of Section 66A and of Section 118(d) of the Kerala Police Act, the validity of Section 69A and the Blocking Rules, and the validity of Section 79 and the Intermediary Guidelines subject to the reading down of Section 79(3)(b) and Rule 3(4).

Q: What does void ab initio mean for Section 66A?

Void ab initio means void from the beginning. A post constitutional statute that violates a fundamental right is, under Article 13(2), void to the extent of the contravention, and the established understanding of the declaration in Shreya Singhal is that Section 66A is treated as though it had never been enacted. The consequences are categorical rather than prospective: no fresh case could be registered under the provision after 24 March 2015 because no offence existed, no pending prosecution under it could lawfully continue, and no conviction recorded under it could stand. The expression is the legal foundation for the Supreme Court’s later directions that the reference to the provision be deleted from every pending case.

Q: What is the chilling effect doctrine in free speech law?

The chilling effect describes the deterrence of lawful speech by a law whose prohibitions are uncertain. A speaker who cannot tell in advance whether a sentence will constitute an offence insures against prosecution by silence, and speech that would never have been prosecuted is lost along with the speech the law could validly reach. In Shreya Singhal the doctrine did concrete work: because the expressions in Section 66A were undefined, the Court held that the provision was liable to be used in a way that would chill protected discussion and advocacy, and counted that effect among the reasons the provision could not stand. The doctrine has since become a standard step in Indian challenges to speech restrictions.

Q: Which provision of the Kerala Police Act did Shreya Singhal strike down?

Section 118(d) of the Kerala Police Act, 2011. That provision punished a person who caused annoyance to any person, or a nuisance, through a communication device or by any other means. The Court struck it down as violative of Article 19(1)(a) and not saved under Article 19(2), for the same reasons that destroyed Section 66A: its operative concept was an undefined reaction, annoyance, and it could not be fitted within any head of permissible restriction. The holding matters beyond Kerala because it shows that the discipline of the judgment binds state legislatures as it binds Parliament, and applies to a police statute as it applies to a technology statute.

Q: What can a person do if a case is registered against them under Section 66A?

The legal position is settled and can be stated in a sentence: no offence under Section 66A exists, and the Supreme Court’s directions of 12 October 2022 require the reference to the provision to stand deleted from any pending proceeding and forbid the registration of any fresh complaint under it. A person facing such a case can place the Shreya Singhal judgment and the 2022 directions before the investigating officer, the magistrate or the High Court, as the stage of the proceeding requires. Which forum and what sequence depend on the individual facts, and this general account is information about the law rather than advice on a case; advice on the specific proceeding should be taken from a lawyer without delay.

Q: Which petitions were heard together with Shreya Singhal?

Ten writ petitions were heard and disposed of together, with Shreya Singhal’s petition, Writ Petition (Criminal) No. 167 of 2012, as the lead case. The batch included the petition of the cartoonist Aseem Trivedi, who had been arrested over political cartoons; the petition of Mouthshut.com, a consumer review platform that carried intermediary obligations; and petitions by the People’s Union for Civil Liberties, by Common Cause, and by the Internet and Mobile Association of India. The spread of petitioners, an individual speaker, an arrested cartoonist, a platform and civil liberties organisations, is part of why the judgment addresses the offence, the blocking power and intermediary liability within a single set of reasons.

Q: What did Shreya Singhal say about the sedition law and Kedar Nath Singh?

The judgment treated Kedar Nath Singh v. State of Bihar, which in 1962 upheld the sedition provision in Section 124A of the Penal Code, as settled law, and used it as the model of a speech provision saved by construction. In Kedar Nath Singh the Court had confined sedition to acts involving incitement to violence or an intention or tendency to create public disorder. Shreya Singhal reasoned from that position: incitement marks the boundary at which Article 19(2) becomes available, and the reason Section 66A could not receive the same rescue was that it contained no equivalent core to confine, since its offence was composed of undefined reactions rather than of a defined act with an established limiting construction.

Q: How did the Shreya Singhal judgment use the doctrine of vagueness?

The Court treated vagueness as a constitutional defect in a penal speech law, drawing on the American void for vagueness authorities and on the Indian decisions, including Kartar Singh v. State of Punjab, that accept the doctrine where a law fails to define an offence with sufficient definiteness. A provision is vague, on this approach, when persons of ordinary intelligence must guess at its meaning and differ as to its application, and when its indefiniteness invites arbitrary and discriminatory enforcement. Section 66A failed the standard completely: none of its operative words is defined anywhere in the Act, and the words point in different directions, to the character of the message, the purpose of the sender and the feelings of the recipient.

Q: What did the Zombie Tracker find about prosecutions under Section 66A?

The Zombie Tracker, a project of the Internet Freedom Foundation built with CivicDataLab and launched on 15 January 2021, assembled district court records from the eCourts platform across eleven States. It found that in those States 229 cases under Section 66A were pending when the provision was struck down in March 2015, that police lodged 1,307 fresh cases under the provision after it had ceased to exist, and that 745 of those cases were still pending and active on 10 March 2021, Maharashtra accounting for 381 and Uttar Pradesh for 245. The project itself records the limits of the dataset: coverage of eleven States and uneven digitisation mean the counted figures understate the all India position.

Q: What did the Attorney General tell the Supreme Court about Section 66A in July 2021?

At the hearing on 5 July 2021 of the application by the People’s Union for Civil Liberties, Attorney General K. K. Venugopal, appearing for the Union of India, explained why the struck down provision was still being used. Section 66A, he submitted, still stands in the Information Technology Act, and the fact that the Supreme Court has struck it down appears only in a footnote; when a police officer has to register a case, the officer reads the section and registers the case without reading the footnote. His suggested repair was to place a bracket against the section recording that it has been struck down and to reproduce the judgment in the footnote. The bench, presided over by Nariman J., described the state of affairs as terrible and issued notice to the Union.

Q: Can Parliament enact a new law to replace Section 66A?

Yes, and the judgment itself marks the space for such a law. Shreya Singhal did not hold that online abuse, harassment, threats or incitement are beyond legislative reach; it held that they cannot be criminalised by undefined words that also capture protected speech. A successor provision would have to name its harm, define its terms, place itself under a head of Article 19(2), and maintain a proximate connection between the words punished and the harm addressed. Proposals for successor provisions were examined within government, including by the committee chaired by T. K. Vishwanathan, whose recommendations were prepared without public consultation and were not published when made. No replacement provision has been enacted, and Section 66A itself has not been repealed.