Two Supreme Court judgments carry the name Puttaswamy, and public discussion merges them so regularly that the merger has become the first obstacle to understanding either. The judgment of 24 August 2017 was delivered by a bench of nine judges who were asked one question: whether the Constitution protects a fundamental right to privacy at all. The judgment of 26 September 2018 was delivered by a bench of five judges who were asked a different question: whether one statute, the Aadhaar Act, survived the standard the nine judges had laid down. The first decision recognised a right and stated the test by which invasions of it are to be judged. The second applied that test to a biometric identity programme and upheld most of it, striking down or reading down the parts that failed. A reader who cannot say which decision did which of these things will misread every commentary that cites the pair, because the 2017 decision decided no question about Aadhaar and the 2018 decision recognised no new right.

The relationship between the two decisions is not an accident of naming. The privacy question reached a nine-judge bench because it arose inside the Aadhaar litigation itself, and the smaller benches hearing the challenges to the identity programme could not proceed until the Constitution’s position on privacy was settled. The first judgment is therefore a general constitutional decision, and it has since been applied in fields far removed from identity documents, including criminal law, sexuality, reproductive autonomy and restrictions on the internet. The second judgment is an application of the first to a single programme, and its authority is correspondingly narrower, except on one point: its treatment of the Money Bill route, which has been expressly doubted by a later Constitution Bench and referred to a larger bench, where it remains undecided. This article keeps the two decisions apart, states the test precisely, records what the 2018 majority upheld and what it cut down, presents the dissent in its own terms, and closes with the question that is still open.

The reference that produced the privacy judgment

The litigation began in 2012, when Justice K.S. Puttaswamy, a retired judge of the Karnataka High Court, petitioned the Supreme Court to challenge the Aadhaar project, then being built by executive action without a governing statute. His case, Writ Petition (Civil) No. 494 of 2012, and the batch of petitions filed alongside it attacked the project on the ground that compulsory enrolment in a biometric database, and the use of that database to condition access to public services, violated the right to privacy. For that argument to run, privacy had to be a fundamental right. The Union of India’s answer was that it was not.

That answer was not a debating point. In August 2015, in the course of the Aadhaar hearings, the Attorney General for India submitted that the Constitution does not guarantee a fundamental right to privacy, relying on two early decisions of the Court: an eight-judge bench decision of 1954 and a six-judge bench decision of 1962, each of which contained observations denying that privacy enjoyed constitutional protection. Benches of three and five judges cannot overrule decisions of eight and six. The three judges hearing the Aadhaar batch accordingly referred the question whether the Constitution protects a right to privacy to a larger bench, and in July 2017 a bench of nine judges was constituted to answer it. The bench comprised Chief Justice J.S. Khehar and Justices J. Chelameswar, S.A. Bobde, R.K. Agrawal, Rohinton F. Nariman, A.M. Sapre, Dr D.Y. Chandrachud, Sanjay Kishan Kaul and S. Abdul Nazeer.

The nine judges answered the referred question and nothing else. Their judgment of 24 August 2017, reported at (2017) 10 SCC 1, was unanimous in its result: the right to privacy is protected as an intrinsic part of the right to life and personal liberty under Article 21 and as part of the freedoms guaranteed by Part III of the Constitution, and the contrary observations in the decisions of 1954 and 1962 stood overruled. Six opinions were delivered. The plurality opinion was written by Justice Chandrachud for himself, Chief Justice Khehar, Justice Agrawal and Justice Nazeer. Justices Chelameswar, Bobde, Nariman, Sapre and Kaul each wrote separately, concurring in the result while developing the reasoning along different lines. Unanimity on the operative answer, delivered by the largest bench that can realistically be assembled, gave the holding a weight that the separate opinions did not dilute: on the existence of the right, there was nothing left to argue about.

What did the Constitution say about privacy before 2017?

Nothing in terms. The word privacy does not appear in the text of Part III, and the judgments of 1954 and 1962 had treated that silence as significant. What existed before 2017 was a line of smaller-bench decisions that had recognised privacy, sitting beside the two larger-bench denials, and it was that conflict the reference was constituted to resolve.

Ten years of litigation before the two judgments

The two judgments sit at the end of a procedural road that itself explains why the questions arrived in the order they did, and a reader who skips it will wonder why a case filed in 2012 produced its foundational ruling in 2017 and its merits ruling in 2018.

The Aadhaar project began as an executive programme. The Unique Identification Authority of India was created by a notification of the Planning Commission in 2009, and enrolment proceeded under executive authority, without a statute. The first constitutional challenges, of which Justice Puttaswamy’s petition of 2012 was the lead, attacked the programme at that stage: a biometric database of the resident population was being assembled, and its number demanded for services, on the strength of executive instructions alone. In September 2013, an interim order of the Court recorded that no person should suffer for not getting an Aadhaar card despite a circular making it mandatory, and that the number was not to be issued to illegal immigrants. The interim position, repeated in orders that followed, was that Aadhaar was to be voluntary and that no benefit was to be denied for want of it, while the constitutional challenge remained undecided.

In August 2015 the case took the turn that determined everything afterwards. A bench of three judges, confronted with the Union’s submission that there was no fundamental right to privacy at all, referred the batch to a Constitution Bench, and by an order of 11 August 2015 recorded that the use of Aadhaar would in the interim be confined to the public distribution system and to the distribution of cooking gas, and that the number’s possession could not be made a condition of other benefits. In October 2015 a Constitution Bench permitted the voluntary use of Aadhaar in a small number of additional schemes, on the Union’s assurance that no one would be deprived of a benefit for want of enrolment. Meanwhile the legislative ground shifted: in March 2016 Parliament enacted the Aadhaar Act, and enacted it as a Money Bill, converting an executive programme into a statutory one in the middle of its own constitutional challenge. Fresh petitions attacked the Act, the Money Bill certification, and the instruments issued under it, and it was within that enlarged challenge that the privacy reference went to nine judges in 2017 and the merits went to five judges in 2018. The sequence matters to the doctrine: the legality stage of the 2017 test, the existence of a law, was answered for the period before 2016 by the absence of one, which is why the pre-Act years of the programme feature so prominently in the dissent and why Section 59, the Act’s validation of the earlier period, required its own defence in 2018.

The road to the nine-judge bench

To understand what the nine judges did, a reader needs the two decisions they overruled and the line of cases they approved, because the 2017 judgment is as much a settlement of the Court’s own past as it is a statement about privacy.

The first obstacle was M.P. Sharma v. Satish Chandra, decided in 1954 by a bench of eight judges. The case concerned search and seizure under the Companies Act, and the challenge invited the Court to import into Indian law something like the Fourth Amendment to the Constitution of the United States. The Court declined. It held that the Indian Constitution contained no counterpart provision restraining search and seizure on privacy grounds, and observed that a power of search and seizure is in any event an overriding power of the State for the protection of social security. Read for what it decided, M.P. Sharma was a case about searches. Read for one sentence in it, it became authority for the proposition that the Constitution knows no right to privacy, and that sentence was the one the Attorney General relied on sixty years later.

The second obstacle was Kharak Singh v. State of Uttar Pradesh, decided in 1962 by a bench of six judges. A man against whom a history sheet had been opened under the Uttar Pradesh Police Regulations challenged the surveillance regime applied to him, which included secret picketing of his house, tracking of his movements, and domiciliary visits by the police at night. The majority struck down the domiciliary visits but upheld the rest, and held that the right to privacy is not a guaranteed right under the Constitution. Justice Subba Rao, dissenting in part, would have recognised the right. The structure of the case mattered in 2017 for two reasons. The majority’s reasoning rested on the compartmentalised reading of fundamental rights associated with the Court’s earliest case law, under which each article was treated as a self-contained code. And the very provision the majority struck down, the night visit to a man’s home, was struck down on reasoning about liberty that sat awkwardly with the denial of privacy in the same judgment.

Between those two decisions and the reference, a different line of authority grew up in smaller benches, which could distinguish the large-bench cases but could not displace them. In Gobind v. State of Madhya Pradesh in 1975, a bench of three judges dealt with police surveillance regulations and proceeded on the footing that a right to privacy could be located in Article 21, subject to restriction on the ground of a compelling public interest. In R. Rajagopal v. State of Tamil Nadu in 1994, the Court recognised a right to be let alone and worked out its limits against the freedom of the press. In People’s Union for Civil Liberties v. Union of India in 1997, the Court held that telephone tapping invades privacy protected under Article 21 and laid down procedural safeguards to govern it. In Selvi v. State of Karnataka in 2010, the Court held that the involuntary administration of narco-analysis, polygraph examination and brain-mapping violates both the protection against self-incrimination and the mental privacy of the subject. Each of these decisions recognised privacy in a concrete setting. Each was delivered by a bench too small to confront the 1954 and 1962 denials directly. That is why the conflict survived for decades, and why, when the Attorney General stood on the two early cases in 2015, only a bench of nine could answer him.

What happened to the smaller-bench cases that had recognised privacy?

They were approved. The order of the nine-judge bench recorded that the decisions after Kharak Singh which had recognised privacy as protected under Article 21 and Part III lay down the position in law. Gobind, Rajagopal and the telephone-tapping case were not casualties of the reference; they were vindicated by it, and the two larger-bench denials were the authorities overruled.

What the nine judges held

The holding can be stated in four propositions, and each of them carries weight in the litigation that has followed.

The first is location. Privacy is protected as an intrinsic part of the right to life and personal liberty under Article 21, and as part of the freedoms guaranteed by Part III as a whole. The plurality reasoned that privacy is not the content of one article among others but a value that runs across the fundamental rights chapter: it is implicated in the equality guarantee of Article 14, because state action that classifies and profiles touches the conditions of equal citizenship; it is implicated in the freedoms of Article 19, because speech, movement, association and residence all presuppose a sphere the State does not enter without justification; and it is anchored in the life and personal liberty of Article 21, because a life without a protected inner and decisional space is not a life of dignity. This plurality holding, that privacy is not confined to any single article, has practical consequences: an invasion can be attacked on whichever article it touches, and a defence must answer under that article’s own limitation clause.

The second proposition is content. Privacy includes at least three families of interest, and the opinions distribute emphasis among them. There is the privacy of the body: the interest in bodily integrity, in not having one’s body searched, sampled or interfered with at the State’s pleasure. There is decisional autonomy: the interest in making the intimate choices that constitute a life, concerning family, marriage, procreation, the home, sexual orientation and personal association, without the State prescribing them. And there is informational privacy: the interest in controlling what is known about oneself, in an age when the collection, aggregation and analysis of personal data allows the State and private entities alike to construct profiles from fragments each of which may look trivial. Because informational privacy is where the Aadhaar argument would later be fought, the point deserves emphasis: the plurality treated the danger as lying in aggregation and profiling rather than in the sensitivity of any single item of data, and it is the statute that later built a legal regime for personal data that inherited that framing. The plurality also grounded the right in dignity and in the idea that privacy is a natural right, inherent in the person rather than granted by the Constitution, which recognises rather than creates it.

The third proposition is the limit. Privacy is not an absolute right. Like the other freedoms of Part III, it may be restricted, but only by a procedure and on grounds that satisfy the test the judgment laid down, which is the subject of the next section. The recognition of the right therefore decided the threshold question and deliberately did not decide the outcomes of future contests: it supplied the standard and left its application to the cases.

The fourth proposition concerns the Court’s own history. The plurality addressed the Emergency-era decision in ADM Jabalpur v. Shivkant Shukla, in which a Constitution Bench had held that during an Emergency a person could not move a court even against an illegal deprivation of life or liberty. Justice Chandrachud described the judgments in that case as seriously flawed, and the reasoning of the privacy plurality, that life and personal liberty are inalienable and recognitional rather than gifted by the Constitution, was written to remove the foundation on which ADM Jabalpur had rested. The formal order of the nine judges overruled M.P. Sharma and Kharak Singh. The treatment of the Emergency precedent was accomplished in the reasoning, and it matters to the Puttaswamy story because it shows the plurality’s location of rights outside the State’s gift doing substantive work.

One further strand of the 2017 reasoning matters for what follows. The plurality and Justice Kaul both addressed the threat that privacy faces from actors other than the State. The plurality observed that the state’s obligation is not exhausted by abstaining from invasion; the modern informational environment is one in which personal data is held and processed by private entities, and the protection of informational privacy requires a legal framework governing them, a task for legislation informed by the report of the committee the government had constituted under Justice B.N. Srikrishna. Justice Kaul, in his separate opinion, identified the dangers to privacy posed by non-State actors in terms and treated the existence of procedural safeguards against abuse as part of the test an invasion must pass. The seeded questions this article must answer return to that strand, because the question whether the privacy right reaches private parties is among the most frequently misstated parts of the whole subject.

The positive obligation: privacy as something the State must also protect

Buried in the 2017 judgment is a proposition that has grown in importance with each year of the data economy: the State’s duty in respect of privacy is not exhausted by refraining from invading it. Fundamental rights in the classical model restrain the State, and the greater part of the Puttaswamy litigation fits that model, since the invader the petitioners feared was the State itself. But the plurality and Justice Kaul both treated the informational environment as one in which the systematic invasion of privacy is carried on by private actors: by platforms that aggregate behavioural data, by employers and service providers who condition goods on disclosure, and by data brokers whose subjects have never dealt with them. Against invasions of that character, a right that binds only the State is a right that protects against the smaller threat.

The plurality accordingly recorded that the State bears an obligation to put in place a regime that protects informational privacy against non-State actors, and it connected that obligation to the committee the government had appointed under Justice Srikrishna to frame a data protection law. Justice Kaul’s concurrence developed the point with the most precision: he identified the threat from non-State actors as a distinct and growing source of danger to the right, treated the protection of the individual against such actors as part of the constitutional enterprise, and included in his safeguards requirement the existence of a framework within which the individual could control and correct the use of personal information. The proposition that emerges, stated carefully, is that the Constitution requires the State to legislate for the protection of privacy in the private sphere, while leaving the content of that legislation to Parliament; the right against private parties operates, in the main, through the statute the State is obliged to enact rather than as a direct constitutional claim against the private party itself.

That careful statement is the one the 2017 judgment supports, and it is frequently misstated in both directions: privacy is said to bind private parties directly, which the judgment did not hold as a general proposition, or to have nothing to do with private conduct, which the judgment expressly rejected. The later development has moved the law a step further along the road the concurrences marked. When the Constitution Bench in Kaushal Kishor held in 2023 that the rights under Articles 19 and 21 may be enforced against persons other than the State, it converted into enforceable doctrine a trajectory that the privacy opinions of 2017 had begun, and students of the subject should read the two developments together: 2017 established that privacy must be protected against private invasion and that the instrument of protection is legislation; 2023 established that the rights themselves are not, in principle, confined to claims against the State. Between the two lies a set of questions, about remedies, about the forums in which such claims run, and about the collision of one person’s privacy with another’s expression, that the cases have opened and not yet closed.

Privacy inside the freedoms: the Article 19 dimension

The plurality’s insistence that privacy runs through the whole of Part III, and not through Article 21 alone, has a consequence that is underappreciated in accounts which treat the judgment as a case about data. Where an invasion of privacy also burdens one of the Article 19 freedoms, the restriction must be justified under that freedom’s own limitation clauses, and the analysis changes character.

Consider the press. The Rajagopal decision of 1994, which the nine judges approved, had already balanced the privacy of the individual against the freedom of publication, protecting the intimate facts of private life while exposing the public conduct of public officials to scrutiny. Under the plurality’s construction, that balance is no longer a courtesy extended by one article to another: privacy is itself an ingredient of the freedoms, and the citizen who speaks, publishes, assembles or moves does so in reliance on a sphere in which the State and the stranger may not follow without justification. Surveillance that falls short of any direct prohibition can therefore burden the freedoms in fact, by the chill it casts on speech and association, and the plurality’s reasoning supplies the terms in which such a burden is argued: the invasion is located, the freedom it burdens is named, and the State is put to the test under whichever articles the invasion touches. The same structure runs through the association and movement cases, through the protection of the anonymous pamphleteer and the confidential source, and through the later internet cases in which the proportionality inquiry was applied to restrictions that were challenged as burdens on speech and trade exercised online. A restriction defended under Article 21’s procedure alone, without reference to the freedom it burdens, is after Puttaswamy an incompletely defended restriction.

This dimension also disciplines the State’s favourite answer to privacy claims, that the interest asserted is not privacy at all but a disguised claim to commit wrongdoing unseen. The plurality’s constellation treats the sphere as protected before its use is examined: the home, the communication and the intimate decision are within the right whether the particular use to which a citizen puts them attracts approval or not, and the State that wishes to enter must justify the entry by the test, not by characterising the occupant. That ordering, protection first and justification second, is the load-bearing premise of the entire judgment, and it is visible in every later case in which a court has refused to let the moral evaluation of the claimant’s conduct substitute for the constitutional inquiry into the State’s power.

The satellite holdings of 2017

Around the central holding, the plurality settled several subsidiary points that have since been quoted as often as the holding itself, and an accurate account should record them.

The plurality rejected the argument that dignity and welfare stand on opposite sides of the case in the way the Union’s submissions had framed it. It had been argued for the State that privacy is an elitist concern, and that in a country where the State struggles to deliver food and housing, the abstraction of informational control should yield. The plurality’s answer was that the Constitution’s commitment to dignity is not divisible by income: the poor have as much claim to a private sphere as the wealthy, the interest in controlling the conditions of one’s life is not a luxury good, and a jurisprudence that treated fundamental rights as the possession of the affluent would misread the chapter in which they sit. That passage of the reasoning matters because it foreclosed a style of argument, the ranking of rights by the perceived needs of their claimants, that had recurred in Indian constitutional litigation for decades.

The plurality also addressed who holds the right. Privacy was located in Article 21, which speaks of persons and not of citizens, and the opinions treat the right as an entitlement of the human person rather than a privilege of citizenship. The point has structural consequences: an Article 19 right can be asserted only by citizens and is subject to that article’s enumerated restrictions, while an invasion of privacy located in Articles 14 and 21 is judged under standards available to every person within the jurisdiction. A number of the plurality’s constructions of the right’s content, the sanctity of the home, the protection of intimate decision, the confidentiality of personal information, draw on this personhood foundation, and it is one reason the judgment’s reasoning travelled so readily into cases about prisoners, patients and non-citizens of the public imagination, claimants who happen to be outside the political community’s favoured circle.

Comparative material runs through the plurality’s reasoning, and the use made of it is worth characterising accurately. The plurality surveyed the treatment of privacy in other constitutional democracies and in the international covenants to which India is a party, not as authority binding on the Court but as evidence that the protection of a private sphere is a recurring feature of constitutional orders that protect dignity at all. The judgment’s own basis remained the Indian text and the Court’s own precedent. Descriptions of Puttaswamy as having imported a foreign right misstate the method: the plurality presented the comparative survey as confirmation of a conclusion it reached from the structure of Part III, from the premise that the fundamental rights are not an exhaustive list of the liberties of a free people but recognitions of them.

Finally, the plurality was careful about what kind of right it was recognising. Privacy, in its analysis, is not a single interest but a constellation: it protects the body against physical invasion, the mind against compelled disclosure of thought and belief, the home against entry and surveillance, the person against the aggregation of their data into a profile, and the individual against the prescription of their intimate choices by the State. The plurality did not attempt to fix the outer boundary of the constellation, and it treated the content of the right as developing through the cases in which particular invasions are challenged. That open texture was a deliberate judicial choice, and it explains a feature of the later case law that puzzles some readers: courts applying Puttaswamy do not ask whether a new claimant’s interest resembles a decided one so much as whether the invasion alleged satisfies the test, with the constellation’s content elaborated case by case.

The six opinions of 2017, and where they differ

Unanimity in the result concealed six distinct pieces of reasoning, and because later benches quote the separate opinions for different propositions, the differences are worth mapping rather than flattening.

The plurality opinion of Justice Chandrachud is the judgment’s centre of gravity. It proceeded historically, tracing privacy through the Court’s own case law and through the constitutional text, and doctrinally, in its rejection of the view that the fundamental rights occupy sealed compartments. That rejection had a pedigree the plurality made explicit: the early decision in A.K. Gopalan v. State of Madras had treated each article as a self-contained code, and the later decisions in R.C. Cooper v. Union of India and Maneka Gandhi v. Union of India had displaced that method, reading Articles 14, 19 and 21 as an interlocking whole, sometimes called in the commentary the golden triangle. The plurality’s location of privacy across Part III rather than inside a single article is the Cooper and Maneka method applied to a new right, and it is why the overruling of Kharak Singh could be accomplished as a restoration of the dominant line rather than as a revolution. The plurality also supplied the judgment’s philosophical register: privacy as the constitutional core of human dignity, as the preservation of personal intimacies and the sanctity of family life, marriage, procreation and the home, and as a right that inheres in the person and is recognised, not conferred, by the Constitution.

Justice Chelameswar’s concurrence approached the right through the structure of freedoms, treating privacy as the organised protection of the liberties the text enumerates and identifying its three constitutive elements in his analysis: repose, sanctuary and intimate decision. Justice Bobde’s concurrence grounded privacy in the very idea of a fundamental right, arguing that privacy is the necessary, indeed inevitable, concomitant of the rights to life and liberty, inseparable from the exercise of every enumerated freedom. Justice Nariman’s concurrence organised the interests protected into three zones he took from the case law and commentary: the intimate zone of the person and the home, the private zone of choice and association, and the zone of informational control, and he was explicit that the right operates against the State subject to the tests each article supplies. Justice Sapre’s concurrence rested the right in the dignity and autonomy of the individual and in the constitutional promise of a just social order. Justice Kaul’s concurrence is the one with the most operational content for later cases: it identified informational privacy as the facet under greatest pressure from technology, addressed the threat from non-State actors directly, insisted that the State’s obligations include the creation of a regime protecting data against private exploitation, and added the procedural safeguards requirement to the conditions of a valid invasion.

The differences among the opinions are differences of route, not of destination, but they have consequences for citation. A later bench needing the test quotes the plurality. A later bench needing the horizontal dimension of privacy, its application to the conduct of private parties, quotes Justice Kaul. A later bench needing the proposition that a law’s invasion is judged against the whole of Part III and not one article alone quotes the plurality’s structural reasoning. Reading the 2017 judgment as a single undifferentiated text misses how the Court actually built it, and understates how much of its later use was made possible by the separate opinions having developed, in advance, the strands the future cases would require.

The test the nine judges laid down

The durable contribution of the 2017 judgment is not the declaration, which a reader can absorb in a sentence, but the standard of review it constructed. A declaration that privacy is fundamental, without a test, would have left every subsequent bench to improvise the balance between the individual and the State. The plurality instead stated the conditions an invasion of privacy must satisfy, and the conditions are cumulative: failure at any stage is failure altogether.

The first requirement is legality. There must be a law. An invasion of privacy by executive action alone, resting on administrative practice or a government circular, does not reach the later stages of the inquiry, because the Constitution requires that a right of this rank be taken away, if at all, by enacted law. The requirement sounds elementary and does real work: it was central to the 2018 bench’s treatment of the mobile-telephone linking directive, which had been issued as an executive instruction, and it is the stage at which surveillance regimes built on notifications rather than statutes are most exposed.

The second requirement is a legitimate State aim. The law must pursue a purpose the Constitution counts as legitimate. Welfare delivery, the prevention of fraud on the exchequer, national security, the prevention of crime: aims of this character pass the stage. An aim that is itself constitutionally impermissible, or a purpose stated so vaguely that any measure could be fitted under it, does not.

The third requirement is proportionality, and the plurality gave the term a structured content rather than leaving it as a judicial impression. The means the law adopts must be rationally connected to the aim it pursues: there must be a nexus between the measure and the object such that the measure is capable of advancing the object. The means must also be necessary, in the sense that there is no equally effective and less intrusive alternative available; a State that can achieve the same end with a smaller invasion must use the smaller invasion. Underlying the plurality’s formulation is the further discipline of balance: the graver the invasion, the more compelling the justification must be, and a measure whose benefit to the stated aim is marginal and whose reach into private life is extensive will fail even if a rational connection can be constructed for it. This structured review has an ancestry in the Court’s earlier work on restrictions to the freedom of trade and profession, where a four-pronged proportionality inquiry had been articulated in Modern Dental College v. State of Madhya Pradesh in 2016; the privacy plurality restated the discipline for Article 21. It belongs to the same family of analysis as the structured review the Court had applied to restrictions on speech in the Shreya Singhal judgment on Section 66A, where the State was likewise put to showing that a rights restriction fitted its stated object.

The fourth requirement is the one a reader must handle with precision, because it is the point on which summaries of the judgment most often mislead. Justice Kaul, in his concurring opinion, added to the three requirements a fourth: that there must be procedural safeguards against the abuse of the power the law confers. He treated the existence of safeguards as a distinct condition, not a part of proportionality itself. The plurality’s own statement of the test enumerated three requirements, and accounts that present the test as a uniform four-part formula agreed by all nine judges overstate the text. Equally, accounts that stop at three and never mention the safeguards element understate the judgment’s working content, because the 2018 bench’s own application of the standard repeatedly turned on the presence or absence of safeguards: the national-security disclosure provision fell for want of them. The accurate statement is the one this article follows: three requirements laid down by the plurality, with a fourth, procedural safeguards, added by Justice Kaul’s concurrence and treated in practice as part of the operative discipline.

Two features of the test deserve note before the article turns to its application. The first is that the test is addressed to the design of a law and not only to its administration. A statute may fail at the stage of necessity because of what it authorises, however benignly the authority promises to use the power; the 2018 dissent’s argument proceeded exactly on that footing. The second is that the test allocates the burden. It is for the State to show the law, the aim, the connection and the absence of a less intrusive alternative. The citizen is not required to demonstrate that the invasion was pointless; the government is required to demonstrate that it was justified. That allocation is what makes the standard a protection rather than a courtesy.

Why does the 2017 judgment matter beyond Aadhaar?

Because it decided a question of constitutional structure, not a dispute about one programme. It settled that privacy is a fundamental right, overruled the two decisions that had denied it for half a century, located the right across Part III, and fixed the test by which later invasions would be judged, in fields the Aadhaar litigation never touched.

The four requirements read closely

Because the test is applied constantly, the content of each of its stages deserves closer treatment than a summary allows, and the differences between the test as a formula and the test as an instrument are located here.

Legality, the first requirement, asks more than whether some legal instrument can be found. The plurality’s requirement is that the invasion be authorised by law, and the law must be accessible and must possess the qualities the Court’s jurisprudence demands of a rights-restricting measure: it must be validly enacted, it must be in force, and it must govern the invasion in question rather than authorise State action of the relevant kind only by implication. The requirement disposes of a class of modern privacy invasions at the threshold: data collection by administrative scheme, surveillance by executive instruction, and conditioning of services by circular all fail here, before any balancing begins, and the 2018 treatment of the telephone-linking directive is the illustration within this very litigation. It also disciplines the legislature: a government that wishes to build an identity or surveillance system must go to Parliament and take the political cost of a statute, which is itself a safeguard of the kind the framers of Part III intended legislative process to be.

The legitimate aim stage asks whether the objective the law professes is one the State may constitutionally pursue, and the inquiry has a screening function that is easy to underestimate. Aims fail at this stage when they are unconstitutional in themselves, when they are pretextual in the sense that the stated objective is not the measure’s operative purpose, and when they are framed at a level of generality, security, efficiency, good governance, that would justify any measure whatever. The welfare-delivery aim in the Aadhaar case passed because it was specific, funded and constitutionally mandated in character: the delivery of subsidies voted by Parliament out of the Consolidated Fund. The stage also fixes the aim for the later stages: proportionality is measured against the aim as found at this stage, and a State may not, having failed the fit between its measure and its stated aim, substitute a different aim mid-argument. That discipline, aim-fixation, is one of the test’s least noticed and most consequential features in practice.

Proportionality, the third requirement, is where most litigation is won and lost, and its internal structure should be kept distinct. The rational connection inquiry asks whether the measure is capable of advancing the aim: a measure that cannot work fails here, and the inquiry is empirical in character even when courts conduct it on affidavit rather than evidence. The necessity inquiry asks whether an alternative measure, equally effective toward the aim, would invade the right less; it is at this stage that the dissent in 2018 directed its strongest attack, arguing that identity assurance for welfare delivery could be achieved by instruments that did not centralise biometrics, and that the Union had not demonstrated the inadequacy of the alternatives. The balancing inquiry, sometimes called proportionality in the strict sense, weighs the benefit of the measure to the aim against the severity of the invasion: a marginal administrative convenience cannot carry a deep invasion, and a deep invasion requires an aim of corresponding weight and a demonstrated connection. The 2017 plurality stated the first two limbs expressly and the balancing discipline by implication in its insistence that the invasion be proportionate to the need; later benches have stated the test as four questions, legitimate purpose, rational connection, necessity, and balancing, and counsel citing the test should be able to move between the two statements without confusion, since the four-question version and the plurality’s version are the same discipline at different levels of articulation.

The safeguards requirement, Justice Kaul’s addition, asks whether the power the law confers is attended by procedures that make its abuse detectable and correctable: oversight by an authority independent of the operator, notice to the individual affected, an opportunity of hearing before disclosure, a remedy that the individual can set in motion without the State’s permission, and limits on retention that force the periodic destruction of accumulated data. The 2018 judgment’s invalidations cluster here for a reason: the safeguards inquiry converts abstract anxiety about a database into examinable questions about who may open it, on whose word, with what record, and subject to whose review. A legislature drafting a data-intensive statute after Puttaswamy drafts, in effect, to this checklist, and the legislative history of every Indian data statute since carries the marks of drafters answering it.

One question the 2017 judgment left incompletely answered is the intensity of the review the test demands: whether the court is to satisfy itself that the State’s choice was the best available, or only that it was a justified one within a range of reasonable choices. The Aadhaar majority’s own language shows the pull of the second approach: it described the standard it applied in terms of fairness and reasonableness and measured the scheme against a larger public interest rather than against a compelling necessity in the strictest sense. Critics of the judgment locate its deference exactly there. Later benches have varied in intensity with the right and the context in issue, and the honest statement of the position is that Puttaswamy fixed the structure of the inquiry while leaving its temperature to be set case by case, which is why the same test struck down a central bank circular in one case and sustained a biometric welfare architecture in another.

Two judgments, one test

The table below is the artifact this article exists to provide. It sets the four requirements of the standard side by side with the use each side of the 2018 bench made of them, so that a reader can see the same test producing opposite answers on the same statute, which is the single most instructive feature of the whole litigation.

Requirement	What the nine judges laid down	How the 2018 majority applied it	How the 2018 dissent applied it	Where later benches have applied the test
Legality	An invasion of privacy requires enacted law; executive action alone is insufficient	The Aadhaar regime rested on the Act of 2016, so the core programme passed; the mobile-linking directive, resting on an executive circular of 23 March 2017, failed and was quashed	The programme from 2009 to 2016 ran without a law, and that original absence infected the whole architecture in the dissent’s analysis	The internet-restrictions cases applied the legality stage to executive orders, requiring a lawful basis and publication
Legitimate aim	The law must pursue a State aim the Constitution counts as legitimate	Targeted delivery of subsidies and benefits funded from the Consolidated Fund of India, and the prevention of leakage and fraud, were legitimate aims	The dissent accepted that welfare delivery was a legitimate aim in the abstract and directed its attack at the means	The cryptocurrency circular case tested the regulator’s measure against its stated aim of protecting the banking system
Proportionality	The means must be rationally connected to the aim, must be necessary in that no less intrusive alternative is available, and must not impose a burden out of scale with the gain	Minimal demographic and biometric data, authentication answering yes or no without disclosing purpose, and targeting confined to subsidies: proportionate for the core programme, disproportionate for bank accounts and telephones	A centralised biometric database, usable across contexts and capable of aggregation and profiling, was disproportionate in design, and the failure of less intrusive alternatives to be considered was fatal	Read down the penal provision on homosexuality, struck down the adultery provision, and structured the review of internet shutdowns
Procedural safeguards	Justice Kaul’s concurrence added safeguards against abuse as a condition of a valid invasion	The national-security disclosure provision fell because a direction by a senior officer without independent or judicial oversight carried no adequate safeguard; the complaint bar fell because it denied the individual a remedy	The dissent found safeguards wanting across the architecture: authentication agencies, requesting entities and data flows operated without an independent regulator or effective individual control	The surveillance litigation of 2021 proceeded on the footing that safeguards and oversight are part of the constitutional inquiry, not an ornament to it

Read the third and fourth columns against each other and the shape of the 2018 disagreement becomes exact. The majority and the dissent did not disagree about the standard. They disagreed about the facts of the architecture: about how much data the system in truth collected, about whether authentication could be confined to its stated purpose, and about whether the safeguards on paper were safeguards in operation. That is why the standard survived the split in the result, and why the standard, rather than the result, is what later courts have carried forward.

The bench that decided Aadhaar

The second judgment was delivered on 26 September 2018 by a Constitution Bench of five judges: Chief Justice Dipak Misra and Justices A.K. Sikri, A.M. Khanwilkar, Dr D.Y. Chandrachud and Ashok Bhushan. The case is reported as Justice K.S. Puttaswamy (Retd.) v. Union of India, (2019) 1 SCC 1. The hearing had run for thirty-eight days spread across the first half of 2018, one of the longest constitutional hearings in the Court’s history, and the evidentiary record was unusual: the Court heard the Chief Executive Officer of the Unique Identification Authority of India make a technical presentation on the architecture of the system, alongside senior counsel attacking and defending it.

Three opinions were delivered. Justice Sikri wrote the majority opinion for himself, Chief Justice Misra and Justice Khanwilkar. Justice Bhushan wrote a separate opinion concurring with the outcome. Justice Chandrachud dissented alone. The result was therefore four to one for the validity of the Act, but the internal structure matters more than the fraction. The propositions that carry the authority of the Court are those of Justice Sikri’s opinion as supplemented by the concurrence; the dissent carries no authority as law and considerable authority as argument, a distinction the later Money Bill litigation would make concrete. The statute whose provisions the bench was testing is the Aadhaar Act as enacted in 2016, which owns the section-by-section account in this series; this article takes from it only the provisions the judgment engaged.

The questions before the bench were correspondingly distinct from the reference. The 2017 bench had been asked whether privacy is a fundamental right. The 2018 bench was asked whether this statute, in whole or in part, invaded that right unjustifiably; whether its enactment as a Money Bill was constitutionally competent; whether the requirement of Aadhaar for the Permanent Account Number under the Income Tax Act survived; and whether the surrounding instruments, the bank-account rules and the telephone-linking directive among them, stood or fell with the Act or independently of it.

What the majority heard: the architecture on the record

Because the 2018 split turned on the facts of the system, the technical record deserves a section of its own. The Court’s inquiry was unusual in Indian constitutional litigation for the extent to which it examined engineering: the Unique Identification Authority’s Chief Executive Officer, Dr Ajay Bhushan Pandey, presented the architecture to the bench under questioning, and the petitioners met the presentation with technical criticism of their own.

The system as described to the Court worked as follows. At enrolment, a resident furnished demographic particulars, name, date of birth, gender and address, together with core biometrics: fingerprints, iris scans and a photograph. The data was transmitted in encrypted form to a central repository, the Central Identities Data Repository, where a de-duplication process compared the biometrics against the existing database to ensure that one person could not enrol twice, and a twelve-digit Aadhaar number was issued. Authentication, the process by which a person proves identity in a transaction, was structured as a query: a requesting entity submitted the Aadhaar number with a biometric or a one-time password, and the repository answered yes or no. On the Authority’s account, the authentication record did not include the purpose of the transaction, the identity of the requesting entity was itself subject to regulation and audit, registered devices and encryption standards governed the endpoints, and the repository was not a record of transactions: it could say that a person had authenticated, at a time, through a channel, but not what the person had bought, drawn or received.

The petitioners’ technical case attacked the adequacy of that description at four points. De-duplication and authentication alike are probabilistic: biometrics match within tolerances, and a system at population scale produces false matches and false rejections in numbers that are small in percentage terms and large in human terms. Biometric identifiers, once compromised, cannot be reissued, so the central store converted a single breach into a permanent exposure. The authentication logs, even without stated purposes, accumulate into a record of a person’s pattern of life, and the silence of a database about purpose was a property of policy and design settings that could change without any change in the system’s capability. And the enrolment ecosystem, agents, enrolment agencies, requesting entities and authentication service agencies, multiplied the points at which data entered hands other than the Authority’s. It was this record, one architecture described by its operator as minimal and siloed and by its critics as centralised and retentive, that the majority and the dissent each evaluated against the same proportionality standard, which is why their conclusions about the architecture, rather than about the standard, carried the case.

What the majority upheld

The centre of the majority’s reasoning is Section 7 of the Act, the conditionality provision, under which the government may require Aadhaar authentication, or proof of possession of an Aadhaar number, as a condition of receiving a subsidy, benefit or service funded from the Consolidated Fund of India. The majority upheld it, and the route to that conclusion shows the test being applied rather than recited.

On the legitimate aim, the majority accepted the Union’s case that the object was the targeted delivery of public expenditure: ensuring that subsidies reach the persons for whom they are voted, and eliminating the diversion, duplication and ghost beneficiaries that the executive record attributed to the absence of a reliable identity instrument. On proportionality, the majority fastened on design features of the system. The data collected at enrolment was, in its assessment, minimal: demographic particulars and core biometrics, not the transactional life of the individual. Authentication, in the architecture described to the Court, returned a yes or no answer and did not disclose the purpose for which authentication was sought, and the Authority did not store the purpose, location or details of the transaction. On the balancing stage, the majority treated the dignity argument as cutting both ways: the petitioners invoked dignity against the database, and the majority invoked dignity in favour of it, holding that an instrument which gives a documented identity to residents who previously had none, and which secures their access to food, work and pensions, serves the dignity of the marginalised rather than offending it. On exclusion, the majority acknowledged that authentication failure was a real phenomenon, with worn fingerprints and connectivity failures falling hardest on manual workers and the elderly, but treated it as an implementation defect to be corrected within the scheme rather than a ground to invalidate it, pointing to the statutory proviso requiring alternative identification where authentication fails and directing that no benefit be denied for want of Aadhaar. Implementation failure, in the majority’s analysis, is a reason to repair a scheme; it is not, without more, a reason to strike down the statute that authorises it.

The majority upheld the passage of the Act as a Money Bill, and because that holding is the one whose authority has since been unsettled, its reasoning needs precise statement. Article 110 defines a Money Bill as a Bill containing only provisions dealing with the enumerated fiscal matters, among them the appropriation of moneys out of the Consolidated Fund and matters incidental to them, and provides that the Speaker’s decision on the question is final. The majority read the Act’s provisions as, in pith and substance, provisions for expenditure from the Consolidated Fund: Section 7, the operative heart of the Act, conditions subsidies payable out of that Fund, and the remaining architecture, enrolment, authentication, the Authority and its machinery, was incidental to that purpose and fell within the incidental limb of the definition. The Speaker’s certification was held to be subject to judicial review only within narrow limits. The dissent’s attack on this reasoning is stated in its own section below; the point to register here is that the majority did not treat the word “only” in Article 110 as doing the exhaustive work the dissent assigned to it, and it was exactly that treatment that a later bench would single out.

The majority also upheld Section 139AA of the Income Tax Act, which makes the quoting of Aadhaar mandatory in applications for a Permanent Account Number and in returns of income. The provision had already survived challenge in separate litigation, and the 2018 bench applied the threefold test to it directly: a law existed, the aim of preventing duplicate and false tax identities was legitimate, and the connection between a unique identity number and that aim was rational and proportionate. Section 59 of the Act, which validated actions taken under the pre-statute executive regime, was upheld as well, against the argument that the years of enrolment before 2016 had been unconstitutional from the start. The Act as a whole, in short, survived: the majority’s invalidations were surgical, directed at specific provisions and instruments, and the surgical quality of the outcome is what public summaries, in both directions of exaggeration, most often lose.

Section 7 confined: what the upheld provision actually authorises

Because Section 7 survived, the most consequential question the judgment answers about it is not whether it is valid but what, as upheld, it permits. The majority read the provision within its own limits, and the reading is narrower than the administrative practice that had grown up around the Act before the judgment.

The condition Section 7 authorises attaches to a subsidy, benefit or service for which expenditure is incurred from the Consolidated Fund of India, or from the consolidated fund of a State where the provision is applied at the State level. The majority’s construction tied the requirement to that fiscal source: a demand for Aadhaar as the condition of a benefit that is not funded from the Consolidated Fund does not draw authority from Section 7, whatever administrative convenience it might claim. That construction is what disposed of the school admission and examination cases within the judgment’s logic, since elementary education is a constitutional entitlement rather than a Consolidated Fund subsidy and the examining bodies’ fees and functions are not the delivery of voted subsidies. It also disciplines the tail of notifications issued under the section during the years of litigation, by which Aadhaar had been demanded across a widening range of schemes: a notification whose benefit lacks the fiscal character the section requires is, on the majority’s reading, outside the section and cannot be saved by it. The proviso the Court emphasised belongs to the same construction: where an individual’s authentication fails, for reasons of biometric mismatch or otherwise, the statute itself requires that alternate and viable means of identification be offered, and the majority directed the authorities to ensure that no person is denied a subsidy, benefit or service for want of Aadhaar or for failure of authentication. Section 7 as it emerges from the judgment is therefore a conditional and provisioned power, not the general power to condition public life on a number that its critics described and that its administrators had at times exercised. The distance between the provision as enacted and the provision as construed is, once again, the measure of the difference between describing this judgment as having upheld Aadhaar and describing what it actually did.

The protective provisions the majority weighed

The majority’s proportionality assessment did not treat the Act as a bare authorisation to collect data. It weighed, on the safeguards side of the balance, the protections the statute itself contains, and a reader assessing the judgment’s reasoning should know what they were, since the dissent’s contrary assessment of the same provisions is what makes the safeguards debate concrete rather than rhetorical.

The Act confines the use of identity information. Its sharing provisions prohibit the sharing of core biometric information in any circumstances and restrict the disclosure of identity information to the channels the statute itself creates, subject to the consent and court-order machinery the judgment examined. The Authority is placed under statutory duties of security and confidentiality, and the Act creates offences, punishable with imprisonment, for unauthorised access to the central repository, for unauthorised disclosure of information collected in enrolment or authentication, and for impersonation at enrolment or authentication. The majority treated this offence architecture as material to the safeguards inquiry: a system in which misuse is criminalised and in which the collected biometrics cannot lawfully be traded or broadcast is, on its reasoning, a different proportionality problem from a system in which collection is unaccompanied by any such floor of prohibition. The dissent’s response was directed at the enforcement structure rather than the prohibitions’ text: the offences depended, under Section 47 as enacted, on the Authority itself choosing to complain, the regulator and the operator were the same body, and a safeguard that activates only at the discretion of the institution it restrains is, in the dissent’s assessment, no safeguard at the point where it matters. The majority’s partial striking down of Section 47 is best read as its partial acceptance of exactly that argument: the prohibitions survived the test, and the gatekeeping of their enforcement did not.

What the majority struck down and read down

The provisions that failed did so at identifiable stages of the test, and the pattern is more instructive than any single item on the list.

Section 57 was the largest casualty. As enacted, it provided that nothing in the Act prevented the use of the Aadhaar number, subject to authentication, for establishing the identity of an individual for any purpose, by the State or any body corporate or person, pursuant to law or contract. The majority struck down the portion of the section that enabled a body corporate or an individual to seek authentication. The reasoning operated at several stages at once: private entities demanding biometric authentication as a condition of commercial services stood outside the subsidy-delivery aim that had justified the programme; a provision whose operative words were “for any purpose” and “pursuant to any contract” imposed no purpose limitation of its own; and the prospect of authentication by an uncontrolled population of private parties was the point at which the architecture of a welfare instrument became, in the majority’s assessment, an instrument of general identification that the proportionality balance would not bear. The striking down was partial, and the precision matters: what fell was the entitlement of bodies corporate and persons to require Aadhaar authentication, not every voluntary use of the number.

Section 33(2) was struck down outright. That provision permitted disclosure of identity information and authentication records in the interest of national security, on the direction of an officer not below the rank of Joint Secretary to the Government of India. The majority held the provision invalid for the absence of adequate safeguards: the decision to release a person’s identity information and authentication records on national-security grounds was left to an officer of the executive, without judicial oversight or any hearing for the individual affected, and a power of that sensitivity carried no procedural protection proportionate to its reach. Section 33(1), which permits disclosure pursuant to an order of a court, was read down in the adjoining holding to require that the individual whose information is sought to be disclosed be afforded an opportunity of hearing. The pair of holdings is the clearest instance in the judgment of Justice Kaul’s fourth requirement doing decisive work: the national-security aim was not questioned, and legality was satisfied by the section itself; what the provision lacked was safeguards, and for that it fell.

Section 47 fell in part. As enacted, it barred courts from taking cognizance of offences under the Act except on a complaint made by the Authority or an officer authorised by it. The majority held the exclusion of the individual victim from the power to set the criminal law in motion to be arbitrary: a person whose identity information was misused or disclosed could not complain, and the Authority, whose own conduct might be in issue, was made the gatekeeper of complaints against the regime it administered. The bar was accordingly struck down to the extent it excluded a complaint by the individual aggrieved, a holding that converted a closed enforcement scheme into one the citizen could invoke.

The metadata holdings operated on the Act and its regulations together. The definition provisions of the Act, read with the authentication regulations, permitted the recording of the transaction context of each authentication. The majority held that the retention of metadata relating to a transaction was impermissible in the form the regulations framed it, on the ground that a record of where, when and for what class of purpose a person authenticated, accumulated over years, is exactly the raw material of the profiling the 2017 judgment had identified as the central informational danger: the danger lay not in any single entry but in the pattern the entries formed. Regulation 26 of the Aadhaar (Authentication) Regulations, 2016, which required the retention of such transaction records, was held impermissible in its framed form. Regulation 27(1), which permitted authentication records to be archived for five years, was struck down in its long-retention aspect: the majority held that authentication records could not be retained beyond six months. A retention period measured in years was, on the majority’s reasoning, out of scale with any audit or grievance purpose that could justify keeping the record at all.

The instruments surrounding the Act fared worse than the Act itself. The directive of the Department of Telecommunications dated 23 March 2017, which had required the re-verification of mobile connections by linking them to Aadhaar, was quashed; it rested on executive instruction rather than law, and failed at the first stage of the test. The amendment to the Prevention of Money Laundering (Maintenance of Records) Rules, which had made the furnishing of Aadhaar mandatory for opening and operating bank accounts, was struck down as disproportionate: the majority applied the balancing stage and held that conditioning every bank account in the country on biometric authentication could not be justified by the anti-money-laundering aim in the form the rule imposed it, and the rule could not shelter under Section 57 after the part of Section 57 that enabled private authentication had itself fallen. The majority also held that Aadhaar could not be made mandatory for admission of children to school, since elementary education is an entitlement and not a subsidy, benefit or service within Section 7, and that bodies conducting examinations, the Central Board of Secondary Education, the National Eligibility cum Entrance Test and the University Grants Commission among them, could not require it. No child, the majority directed, was to be denied a benefit or service for want of an Aadhaar number. The general position described in this article states what the Court held about the instruments before it; it does not substitute for advice on an individual case, and anyone contesting a demand for Aadhaar in a specific transaction should take advice, since forum rules and limitation periods are unforgiving.

The concurrence, and the question of exclusion

Justice Bhushan’s separate opinion repays attention because it shows a second route to the majority’s result, and because its points of difference map the ground on which a differently constituted bench might have moved. He agreed that the Act passed the threefold test and upheld Section 7, Section 59 and Section 139AA. He agreed that the part of Section 57 enabling private authentication could not stand. He approached exclusion with an evidentiary insistence the lead opinion shared but stated more sharply: the implementation of a scheme, in his analysis, could not be judged unconstitutional for defects in its working without material on record demonstrating that exclusion had resulted from the scheme’s design, and the authorities were under a duty to take steps to minimise whatever exclusion occurred and to put alternative mechanisms in place. The concurrence and the majority together therefore rested the disposition of the exclusion argument on a distinction between a scheme that excludes by design and a scheme that excludes in operation, with the Constitution, on their reading, condemning the first and requiring repair of the second.

The factual dispute beneath that distinction was the most humanly concrete part of the case. The petitioners’ record included affidavits and field studies describing beneficiaries denied rations, pensions and wages because authentication failed: fingerprints worn smooth by manual labour, iris scans defeated by cataracts, connectivity absent at the point of sale, seeding errors that matched a person’s entitlements to another’s number. The Union’s record answered with enrolment and authentication statistics at national scale and with the proviso structure of the Act, under which a person whose authentication failed was to be provided alternate means of identification. The majority’s treatment, taking the statistics and the proviso together, held that exclusion was neither the scheme’s purpose nor a demonstrated consequence of its design at the level that would justify invalidation, while directing that the state ensure no one was denied a benefit for want of Aadhaar. The dissent treated the same material as the predictable output of a probabilistic technology made compulsory for subsistence goods, and therefore as a design fact. The division between the opinions on exclusion is thus the division on proportionality itself, expressed in evidence rather than doctrine, and it is the aspect of the case that later commentary, critical and supportive alike, has returned to most often.

The dissent, in its own terms

Justice Chandrachud’s dissent was a lone opinion, and it disagreed with the majority at the level of design rather than detail. Where the majority asked whether the Act’s provisions could be justified one by one against the test, the dissent asked whether the architecture itself, a centralised biometric database built from 2009 and offered as the key to an expanding range of transactions, could survive the test at all. Its conclusion was that the entire Aadhaar programme suffered from constitutional infirmities, that the Act was unconstitutional in its entirety, and that the data already gathered had been gathered under a regime that failed the Constitution.

The Money Bill limb of the dissent is the part that has travelled furthest. Article 110, the dissent reasoned, permits a Bill to be treated as a Money Bill if it contains only provisions dealing with the enumerated matters, and the word “only” is the operative discipline: it confines the special procedure, under which the Council of States may only recommend and cannot amend or reject, to Bills that are in truth confined to finance. The Aadhaar Bill contained substantial provisions that were not provisions about taxation, borrowing, appropriation or the Consolidated Fund: it established a statutory authority, created an identity regime for every resident, conferred authentication powers, and created offences. Those provisions were not incidental to expenditure; they were a free-standing regulatory scheme attached to a subsidy condition. To certify such a Bill as a Money Bill was, in the dissent’s phrase, a fraud on the Constitution: it debated and enacted a general law while bypassing the bicameral process the Constitution prescribes, depriving the Council of States of the role the Constitution assigns it. The dissent also held that the Speaker’s decision under Article 110(3) is subject to judicial review, since a certification that violates a constitutional limitation is not protected by the finality the provision confers. The majority’s use of pith and substance reasoning, borrowed from the law of legislative competence between List entries, was in the dissent’s view misplaced: that doctrine resolves overlaps between two legislatures’ fields, and has no application to a definition that turns on the word “only”.

The proportionality limb mirrored the majority’s test and inverted its factual assessment. Authentication architecture, the dissent held, was not the minimal, confined mechanism the majority described. Biometric identifiers are immutable: a password can be changed, a fingerprint cannot, and a compromised biometric is compromised for life. A central database that links a resident’s identity across databases creates the capability of profiling and, in combination with authentication records retained across contexts, of tracking the transactions of daily life, and a court applying proportionality must judge the capability a system creates and not only the restraint its operators promise. Less intrusive alternatives to biometric authentication for establishing identity and eliminating duplicates had not been shown to be inadequate, so the necessity stage failed. Exclusion was treated not as an implementation blemish but as evidence about the design: a system whose authentication failures fall predictably on the elderly and on manual workers, for a programme whose justification is the delivery of food and work, carries its disproportion in its own operation. And on safeguards, the dissent found the structure wanting at every level: requesting entities, authentication agencies and the Authority operated without an independent data regulator, without individual control over the use of one’s own identity information, and with grievance mechanisms the dissent regarded as illusory.

The dissent’s remedial direction shows how complete its conclusion was. Invoking the Court’s power to do complete justice, Justice Chandrachud directed that the data collected under the programme not be destroyed for a period of one year, that it not be used for any purpose in that period, and that at the end of the year, if no fresh legislation conforming to the principles of the judgment had been enacted, the data be destroyed. A majority reader should notice what the dissent did not do. It did not question the 2017 standard; it applied it. It did not hold welfare delivery to be an illegitimate aim; it held the means disproportionate. Its disagreement with the majority was therefore examinable point by point, and later litigation on the Money Bill question has treated its reasoning on Article 110 as the argument to be answered.

How did the majority and the dissent each read Article 110?

In opposite ways. The majority read the Act in pith and substance as a law about expenditure from the Consolidated Fund, with the architecture incidental to it, and upheld the certification. The dissent read the word “only” in Article 110 as an exhaustive confinement, found a regulatory scheme inside the Bill, and held the certification a fraud on the Constitution.

The remedial shape of the majority judgment

The majority’s choice of remedies is itself an exercise of the proportionality discipline, and it deserves notice because it determined what the judgment did to the country the day after it was delivered.

The majority struck down where a provision’s defect was in its grant of power: the private-authentication limb of Section 57 and the national-security disclosure power in Section 33(2) could not be narrowed by interpretation, because the vice was the power’s existence in that form, so they fell. It read down where the provision had a valid core and a defective reach: Section 33(1) survived with the hearing requirement written into its operation, and the definition provisions survived with transaction metadata excluded from what may lawfully be recorded. It struck down in part where a bar was overbroad: Section 47’s exclusion of the individual complainant fell while the offence framework it gated remained. It quashed the subordinate instruments, the telephone directive and the bank-account rule, in full, because instruments of the executive stand or fall as made and neither could be rewritten by the Court. And it left the data already collected under the programme in place, issuing no direction for its destruction and no retrospective invalidation of the transactions the system had processed, while cutting off the retention that would have carried the records forward indefinitely. The pattern is consistent: the remedy in each instance is proportioned to the defect found, the statute’s working core is preserved, and the invalidations take effect as removals of identified powers rather than as a general condemnation. Set against the dissent’s remedy, a freeze on the data followed by destruction if Parliament did not legislate afresh, the difference between the opinions is a difference about the judicial role as much as about Aadhaar: the majority treated itself as a court correcting a statute at the points where it failed the test, and the dissent treated the statute’s failure as total, so that the Court’s duty was to return the subject to Parliament with the slate cleared. Later readers of the case, in particular the benches that have since had to remedy rights violations in data and surveillance matters, have had both remedial models before them, and the choice between surgical correction and total invalidation has become one of the standing questions of Indian public law practice.

What the 2018 bench did not decide

A judgment’s boundaries are part of its meaning, and the 2018 bench left several matters outside its holdings, some deliberately and some because they were not before it.

The bench did not enact, and did not purport to enact, a data protection law. It recorded the need for one and pressed the Union toward the legislation the Srikrishna Committee had framed, but the protection of informational privacy against the generality of processors, public and private, was left to Parliament, and the shape of that statute was a legislative question on which the judgment supplies principles rather than a text. The bench did not decide the validity of the amendments Parliament later made to the Aadhaar Act, including the framework for voluntary offline verification introduced in 2019; those provisions carry their own litigation history and stand or fall on their own terms. The bench did not disturb the position that enrolment in Aadhaar is, in its statutory form, a facility available to residents rather than a proof of citizenship, and it directed that the number not be issued to illegal immigrants in terms that left the administrative design of that exclusion to the authorities. The bench did not review the general law of interception and surveillance, which stands on other statutes and other precedent. And the bench did not finally settle the question with which it had begun: whether the certification of the Act as a Money Bill was correct in law. It answered that question with a holding, and the holding stands, but the reasoning by which it answered has been expressly reserved for reconsideration by a larger bench, so the most accurate description of the Money Bill position after 2018 is that it is decided and doubted at once.

The Money Bill question after 2018

The 2018 holding on the Money Bill issue did not hold the field, and the manner of its unravelling is part of this article’s subject because the brief of any honest account of the Aadhaar judgment must record that its most contested holding has been doubted by the Court itself.

The doubt was expressed in Rojer Mathew v. South Indian Bank Ltd., decided on 13 November 2019 and reported at (2020) 6 SCC 1, in which a Constitution Bench of five judges was considering the tribunal provisions of the Finance Act, 2017. The bench observed that the Aadhaar majority had not substantially discussed the effect of the word “only” in Article 110(1), and had not examined the repercussions of a finding that some provisions of an enactment passed as a Money Bill do not conform to the enumerated clauses. It noted that Justice Bhushan’s concurrence had reasoned in terms of dominant purpose, while the lead majority opinion had not adverted to pith and substance at all, so the interpretive basis of the Aadhaar holding was itself unclear. Being a bench of equal strength to the Aadhaar bench, it could not overrule that holding; it accordingly referred the interpretation of Article 110 to a larger bench of seven judges, and the dedicated comparison of Money Bills and Finance Bills is where the state of that reference belongs in this series. The reference remains pending and undecided: the seven-judge bench has not delivered a decision, and chief justices have on more than one occasion indicated an intention to constitute it. Until it is answered, the Aadhaar majority’s reading of Article 110 stands as the operative law while standing under an express, recorded doubt from a bench of coordinate authority.

Two further proceedings mark the state of the question. Review petitions against the 2018 judgment were dismissed in January 2021 by four judges to one; the dissent on the review bench, again by Justice Chandrachud, held that it was a constitutional error to close the matter while the larger-bench reference on the Money Bill question was outstanding, and that the reviews should have awaited it. And in the challenge to the amendments made to the money-laundering statute through successive Finance Acts, the bench that decided the PMLA appeals left the Money Bill validity of those amendments open, expressly, for the larger bench: the same question, arising on a second statute, was held in suspense by the same unresolved reference. A holding that began as one limb of one judgment has thus become the outstanding procedural question of Indian legislative law, and the Aadhaar Act continues to operate under a certification whose supporting reasoning its own court has flagged for reconsideration.

What did the 2018 bench direct about a data protection law?

It pressed the Union to enact a robust data protection regime, drawing on the Srikrishna committee’s report with such modifications as the government judged appropriate. The direction carried the 2017 plurality’s reasoning into practice: informational privacy, against the State and against private processors alike, was to be protected by statute, not left to scheme-by-scheme architecture.

The test at work in later cases

The measure of the 2017 judgment is the use later benches have made of it, and the record is dense enough to state as a sequence rather than a sample.

Within a fortnight of the Aadhaar verdict, the test and the holding together decided the largest questions of personal liberty then before the Court. In Navtej Singh Johar v. Union of India, decided on 6 September 2018, a Constitution Bench read down Section 377 of the Indian Penal Code so that it no longer criminalised consensual sexual relations between adults in private. The opinions drew directly on the privacy judgment: on decisional autonomy as a protected facet of privacy, on sexual orientation as an attribute of identity and dignity, and on the proposition, which the privacy plurality had established by overruling the older cases, that fundamental rights are not held at the sufferance of majoritarian approval. In Joseph Shine v. Union of India, decided on 27 September 2018, the Court struck down the adultery offence, resting in part on the autonomy and dignity reasoning of the privacy decision: a provision that treated a married woman as the property of her husband could not survive a Constitution under which her choices were her own. In the Sabarimala case, Indian Young Lawyers Association v. State of Kerala, decided on 28 September 2018, the opinions engaging the exclusion of women from the temple drew on the same stock of dignity and autonomy reasoning, though the privacy analysis there operated alongside the articles on religion and equality rather than at the centre. In Shafin Jahan v. Asokan K.M., the Court had already, earlier in 2018, set aside a High Court order annulling a marriage, grounding the choice of a partner in the liberty and privacy the nine judges had recognised: the choice of whom to marry, the Court held, lies within the core of personal liberty, and neither family nor State may dictate it.

The proportionality limb travelled further still, because it gave courts a structured way to review State action that did not depend on the subject matter being intimate. In Anuradha Bhasin v. Union of India, decided on 10 January 2020, the Court reviewed the restrictions imposed in Jammu and Kashmir after the constitutional changes of August 2019, held that the freedoms of speech and of trade exercised through the internet are protected under Article 19, and required that restrictions on them be justified under a proportionality inquiry, with orders published and subjected to review; indefinite suspension was held impermissible. In Internet and Mobile Association of India v. Reserve Bank of India, decided on 4 March 2020, the Court set aside a Reserve Bank circular that had cut off banking services to cryptocurrency businesses, applying proportionality to economic regulation and holding that the regulator had not shown the measure to be proportionate to any demonstrated harm. The surveillance questions that the Aadhaar architecture had raised returned in Manohar Lal Sharma v. Union of India, the Pegasus proceedings, in which the Court, by an order of 27 October 2021, constituted a committee of experts under the supervision of a retired judge of the Court, reasoning that the State does not get a free pass every time the spectre of national security is raised, and that the unauthorised surveillance of citizens implicates the privacy the nine judges had recognised. In X v. Principal Secretary, Health and Family Welfare Department, decided on 29 September 2022, the Court held that the distinction between married and unmarried women in the medical termination of pregnancy legislation could not stand, grounding reproductive choice in the decisional autonomy and bodily integrity that stand at the centre of the privacy holding. And in Kaushal Kishor v. State of Uttar Pradesh, decided on 3 January 2023, a Constitution Bench held by four to one that the rights under Articles 19 and 21 are enforceable not only against the State but also against persons other than the State, extending into enforceability the line of reasoning that in 2017 had begun with the recognition that privacy is invaded by private actors as well as public ones.

Two patterns in that sequence are worth stating plainly. The privacy holding migrated from its origin: a case about a database produced the authority by which consensual intimacy was decriminalised and reproductive choice protected, outcomes no party to the Aadhaar litigation was principally arguing about. And the proportionality test detached itself from privacy altogether and became a general discipline of justification, applied to internet shutdowns, financial regulation and surveillance with the same stages the plurality had enumerated. Neither pattern was an accident of citation. Both follow from the 2017 bench’s choice to state a general right and a structured test rather than to resolve the database dispute on narrow grounds.

Why the two-stage structure mattered

The litigation’s strangest feature, that the right was declared by nine judges who decided nothing about the programme that provoked the declaration, and the programme was judged by five judges who had not framed the right, was not a procedural curiosity. It shaped both judgments, and understanding it explains both their strengths and the criticisms each attracts.

The advantage of the structure fell to the right. A bench deciding privacy in the shadow of a concrete and politically charged database might have been tempted to calibrate the right to the case: to find a privacy interest strong enough to discipline Aadhaar and no stronger, or weak enough to let the programme pass. The reference procedure removed that temptation. The nine judges addressed the constitutional question in the abstract, against the whole run of precedent from 1954 forward, with no evidentiary record about authentication failure rates or database architecture before them. The right they declared was therefore framed at the level of principle a constitutional text demands, and its content, dignity, autonomy, informational control, could be carried into later cases about sexuality or reproduction without the distortion of having been sized to fit an identity scheme. The generality that critics of the 2017 judgment sometimes note, its willingness to speak in the language of values, is the signature of a bench answering a referred question, and it is the quality that let the judgment become the foundation it became.

The cost of the structure fell on the application. The five judges of 2018 inherited a test stated without reference to the technology they had to judge. Necessity, the inquiry whether a less intrusive alternative would serve, means one thing for a search of a house and another for a population-scale biometric system, where the alternatives are other architectures with their own failure modes, costs and exclusions, and where the comparison cannot be run as an experiment. The majority resolved the difficulty by deference on the engineering facts; the dissent resolved it by judging the capability of the design rather than its described operation. Neither resolution is dictated by the test itself, and the divergence between them is the residue the two-stage structure left behind. It is also the reason the standard has proved more portable than the verdict: later benches applying the test to shutdown orders, banking circulars and surveillance have set its intensity afresh in each context, treating Puttaswamy as the statement of a method whose calibration belongs to the case in hand. A reader who grasps the two-stage structure can predict, for any future application, where the argument will be fought: rarely on the test’s existence, often on the facts it is fed, and always on the temperature at which it is applied.

The criticisms, the defences, and what the record supports

The stance of this series on contested material is to state each case at full strength and to adopt neither, and the Aadhaar judgment attracts three criticisms serious enough to require statement in those terms.

The first criticism concerns the proportionality inquiry itself. The constitutional scholar Gautam Bhatia, who has written extensively on both judgments, has argued that the majority conducted the proportionality analysis at a high level of generality: that it accepted the architecture of the system substantially as the Union and the Authority described it, evaluated the programme against its stated purpose at the level of the programme as a whole, and did not press the necessity stage, the stage that asks whether a less intrusive measure would achieve the same end, with the particularity the test as laid down in 2017 appeared to demand. On this criticism, the gap between the two judgments is a gap of application: the standard of 2017 was exacting, and the application of 2018 was deferential, and the difference is located not in the doctrine but in the Court’s willingness to interrogate the State’s factual premises. The defence of the majority, as it emerges from the judgment’s own reasoning, is institutional. A court reviewing a legislative and administrative choice about the delivery of benefits across a population of the size of India’s is not the body equipped to design the alternative identity system, to weigh authentication technologies against one another, or to substitute its engineering judgment for that of the legislature and the executing authority. Proportionality, on the defence, disciplines the court to test the measure against its aim and to strike down what exceeds it, as the majority demonstrably did with the bank rules, the telephone directive and the private-use provision; it does not license the court to hold out for the least intrusive system imaginable. Both accounts describe the same judgment, and a reader can test them against the text: the parts of the judgment where the majority did insist on particulars, the metadata and retention holdings above all, sit in visible tension with the generality the critics identify elsewhere, and the honest summary is that the inquiry’s intensity varied across the provisions under review.

The second criticism concerns the Money Bill holding. Critics of the certification, of whom the dissent supplied the judicial form and commentators including Bhatia the academic form, argue that the majority read the definition in Article 110 expansively: that a test of pith and substance, or of dominant purpose, converts a definition built on the word “only” into a test of overall character, under which almost any Bill that touches public expenditure can be routed past the Council of States. The consequence, on this argument, is structural and not confined to Aadhaar: the Finance Acts that followed carried amendments to tribunal law, to political funding and to the money-laundering statute through the same gate, and each of those enactments inherited the protection of the Aadhaar reasoning. The defence, which the majority’s reasoning embodies and which the Union pressed, is that the Constitution itself commits the certification to the Speaker and declares the decision final; that the dominant purpose of the Aadhaar Bill, whose operative provision conditioned expenditure from the Consolidated Fund, lay within the enumerated matters; and that the remaining provisions were machinery for that expenditure and so incidental to it. The record’s own comment on this dispute is the most telling fact available: a Constitution Bench of the Court, in Rojer Mathew, found the majority’s treatment of the word “only” wanting and sent the question to seven judges. A court does not refer what it considers well settled, and it does not overrule by a bench of equal strength what it considers plainly wrong; the reference is the institutional form of a serious, unresolved doubt, and it is the state of the question that any accurate account must leave standing.

The third criticism runs in the opposite direction and should be recorded with equal care. Defenders of the programme, and the majority judgment itself, point out that the judgment’s invalidations, taken together, dismantled much of what had made the identity system objectionable as a general surveillance instrument while preserving the use case, welfare delivery, for which the evidence of benefit was strongest: the private sector’s power to demand biometric authentication was removed, the national-security disclosure route was closed, retention was cut from years to months, transaction metadata was curbed, and the two instruments that had extended Aadhaar into every bank account and telephone in the country were struck down. On this account the judgment is a working demonstration of proportionality doing its proper job, trimming a programme to its justification rather than indulging the binary choice between upholding everything and destroying everything. The counter-observation, pressed by the critics, is that the trimming left the centre intact: a mandatory biometric condition on the receipt of food and work by the poorest residents, administered by the same architecture, and immunised in its passage by a Money Bill holding the Court has since doubted. The reader has both accounts here in the terms their proponents use. What the record does not support is either of the two shorthand versions of the case that dominate public argument: that the Court struck down Aadhaar, which it did not, or that the Court approved Aadhaar without condition, which it also did not.

The legislative sequel: from the committee room to a data protection statute

The 2017 plurality had identified a data protection law as the Constitution’s unfinished business, and the 2018 majority repeated the direction in pointed terms, pressing the Union to enact a robust regime drawing on the report of the committee chaired by Justice B.N. Srikrishna. The legislative history that followed is part of the judgments’ effect, and it belongs in this article in outline. The committee, constituted by the Ministry of Electronics and Information Technology while the privacy reference was under consideration, reported in July 2018 with a draft bill. A revised bill was introduced in Parliament in 2019 and referred to a joint committee, whose deliberations extended its scope in contested directions, and the bill was withdrawn in August 2022. A fresh bill was introduced in 2023 and enacted as the Digital Personal Data Protection Act of that year, a statute whose design, built on consent and notice, on the obligations of data fiduciaries and the rights of data principals, and on exemptions for the State that have themselves attracted constitutional argument, is the legislative generation to which the judgments gave rise. That statute’s text and machinery belong to its own articles in this series; the point for the Puttaswamy account is causal and modest. The nine judges did not write a data protection law, and the five judges could not. But both judgments treated such a law as a constitutional necessity rather than a policy option, and the vocabulary in which the eventual statute is debated, purpose limitation, data minimisation, the fiduciary character of the processor’s power over the principal’s information, entered Indian legislative discourse through the litigation and the committee it provoked. Proportionality supplied the courtroom discipline; the committee and the statute carried the same ideas into the drafting room.

How the two judgments are used in court

For the advocate, the two judgments have settled citation values, and a short map of them is worth more than another summary.

For the existence and location of the right, the 2017 judgment is cited, and cited alone: no later authority is needed to establish that privacy is a fundamental right, and no argument to the contrary is available after a unanimous nine-judge bench. For the test, the plurality’s statement of the three requirements, with Justice Kaul’s safeguards condition, is the citation, supplemented by the four-question articulation in the later proportionality cases where the bench prefers it. For the horizontal dimension, the proposition that privacy is invaded by private actors and that the State bears an obligation to protect it against them, the citation is Justice Kaul’s concurrence and the plurality’s observations on non-State actors, later carried into enforceability by the Constitution Bench in Kaushal Kishor. For the Aadhaar Act itself, the 2018 majority is authority for the validity of Section 7 and of Section 139AA, and for the invalidity of the instruments and provisions it struck down; it is not authority that Aadhaar may be demanded by private entities, by banks under the struck-down rule, or for school admission or examinations, and counsel who cite it for those propositions invite the passages that hold the opposite. For the Money Bill, the 2018 holding stands but is cited with the caveat its own court attached to it: since Rojer Mathew, a careful advocate states the holding, states the doubt, and states the pending reference in the same breath, because a bench that hears only the first of the three will treat the omission as a failure of candour. The dissent, finally, is cited as dissents in Indian constitutional practice are cited: not as law, but as the reasoned position available to a future bench, and on the Money Bill question its reasoning has been absorbed into the doubts of a Constitution Bench, which is the closest a dissent comes to vindication without a formal overruling.

How the two judgments are examined

The brief for this article records that the two judgments are among the most examined material in the law syllabus, and the pattern of examination explains the confusions the article has been at pains to prevent, because examiners frame their questions at exactly the joints where public summaries go wrong.

Questions on the 2017 judgment test the distinction between the holding and its apparatus. The holding, that privacy is a fundamental right intrinsic to Article 21 and spread across Part III, can be stated in a line and earns a line’s worth of credit. The marks sit in the apparatus: in the candidate’s ability to name the two decisions overruled and the bench strengths that made a nine-judge reference necessary; to state the test with its full complement of requirements and to attribute the safeguards element correctly to Justice Kaul’s concurrence rather than to the plurality; to explain why the right’s location across three articles, rather than inside one, changes how a restriction on it is defended; and to identify the smaller-bench cases the reference vindicated. Questions on the 2018 judgment test inventory and attribution in the same way: which provisions were struck down, which were read down, which instruments fell with them, and how the bench divided, not merely in the four-to-one result but in the three opinions that produced it. Questions on the relationship between the judgments are the most discriminating of all, because they expose the candidate who has merged them: the expected answer runs through the reference, the standard, the application, the dissent’s rival application, and the reference that the Money Bill holding itself generated, in that order. A candidate who can also state what remains undecided, the seven-judge question, and can say why a pending reference does not suspend the holding it doubts, has the measure of the whole subject. This article has been organised so that a reader who works through it in sequence holds, in order, every element those questions reward, and the table above is the revision instrument for the centre of it. Candidates who keep their own alignment of the two benches, opinion by opinion, can work on that comparison in an office file on ReportMedic rather than trusting any published summary, this one included.

Six errors that recur in accounts of these judgments

The two judgments generate confident misstatement at a rate unusual even for constitutional law, and correcting the standard errors is a service this article should perform expressly, because each error disables a reader’s ability to use the cases.

The first error is to call the 2018 decision “the Puttaswamy privacy judgment”. The privacy judgment is the 2017 decision of the nine judges, reported at (2017) 10 SCC 1. The 2018 decision is the Aadhaar judgment, reported at (2019) 1 SCC 1, which applied the privacy right to a statute and recognised nothing new. A citation that attributes the right’s recognition to 2018, or the Aadhaar holdings to 2017, has confused the two things this article exists to keep apart, and the confusion is compounded by the case name being identical: the discipline is to cite by year and by bench strength, the nine-judge Puttaswamy or the five-judge Puttaswamy, and never by name alone.

The second error is to state the test as three requirements and stop. The plurality’s test has three limbs, but Justice Kaul’s concurrence added procedural safeguards, the 2018 judgment invalidated provisions on safeguard grounds, and later courts routinely inquire after safeguards as part of the operative discipline. An answer that omits the fourth element will fail precisely on the class of provision, disclosure powers and data retention, that the test has most often invalidated in practice.

The third error is to say that the Court struck down Aadhaar, or that it struck down Section 7. It did neither. The Act survived, the conditionality provision survived, and the tax provision survived; what fell was a list of identified provisions and instruments, set out in this article, and any summary that converts the surgical outcome into a general one misleads in whichever direction it errs.

The fourth error is the mirror image: to say that the Court approved Aadhaar as enacted. It did not. Private entities lost the power to demand authentication, the national-security disclosure route was closed, retention was cut to six months, metadata was curbed, and the two instruments extending Aadhaar into banking and telephony were struck down. The Aadhaar that operates after the judgment is a materially smaller scheme than the Aadhaar that entered the courtroom.

The fifth error is to describe the Money Bill question as settled in favour of the certification. The 2018 majority upheld the certification, and a Constitution Bench in Rojer Mathew thereafter recorded that the majority’s reasoning had not engaged the word “only” in Article 110 and referred the question to a bench of seven judges, where it remains undecided. The accurate statement is that the holding stands and its reasoning is under reference; any statement of the position that omits either half of that sentence is wrong.

The sixth error is to treat the dissent as a footnote. A dissent that has been taken up by a later Constitution Bench as the ground for doubting the majority, that framed the argument on which a seven-judge reference waits, and that predicted the litigation path of the Money Bill issue across a second and third statute, is an operative part of the law’s development even though it binds no one. Readers who skip it will not understand why the question they think was answered in 2018 is still being argued.

Closing verdict

The series has repeatedly found that a judgment’s reputation attaches to its outcome while its effect attaches to a test or a definition, and the Puttaswamy litigation is the constitutional apex of that pattern. The outcome of 2018, that a biometric identity programme survived with its edges cut away, has had a bounded importance: it governs one statute, parts of it are under a recorded doubt, and the programme it saved has continued to be litigated provision by provision. The method of 2017 is the contribution without boundaries. A structured demand that the State show its law, its aim, the fit between its means and its aim, the absence of a gentler means, and the safeguards that discipline its power has been carried into fields the Aadhaar petitioners never argued, and it is the form in which Indian courts interrogate restrictions on liberty of every kind. Majority and dissent in 2018 applied the same standard and reached opposite results, which is ordinarily counted a weakness in a test; in this instance it is better read as the test doing its work of locating the disagreement precisely, in the facts of an architecture and the weight of safeguards, rather than in slogans about privacy or welfare. One question remains genuinely open, and it is not a question about privacy at all. Whether a Bill of the Aadhaar kind may lawfully be carried through Parliament as a Money Bill awaits the seven judges to whom the question stands referred, and until they answer, the final word on the most consequential procedural holding of the 2018 judgment has not been written.

Frequently Asked Questions

Q: What did the nine judge bench decide about privacy?

It decided, unanimously, that the right to privacy is protected as an intrinsic part of the right to life and personal liberty under Article 21 and as part of the freedoms guaranteed by Part III of the Constitution. The bench held that privacy is not confined to any single article but runs across the fundamental rights chapter, that it includes at least bodily integrity, decisional autonomy and informational privacy, and that it is not an absolute right, so an invasion of it must satisfy the requirements of legality, a legitimate State aim and proportionality. The decision is reported at (2017) 10 SCC 1.

Q: Which earlier judgments did Puttaswamy overrule?

It overruled the decision in M.P. Sharma v. Satish Chandra of 1954, in which a bench of eight judges had held against the existence of a constitutionally protected right to privacy in the context of search and seizure, and the decision in Kharak Singh v. State of Uttar Pradesh of 1962, in which a bench of six judges had held that the right to privacy is not a guaranteed fundamental right, each to the extent that it denied constitutional protection to privacy. The bench recorded that the decisions after Kharak Singh which had recognised privacy, including Gobind, Rajagopal and the telephone-tapping case, lay down the correct position in law.

Q: What is the proportionality test for privacy?

It is the standard the 2017 bench laid down for judging an invasion of privacy. The invasion must satisfy three requirements stated by the plurality: legality, meaning there must be an enacted law authorising it; a legitimate State aim, meaning the objective pursued must be one the Constitution counts as legitimate; and proportionality, meaning the means must be rationally connected to the aim and necessary, in the sense that no equally effective and less intrusive alternative is available, with the burden of the invasion weighed against its gain. Justice Kaul’s concurrence added a fourth requirement: procedural safeguards against abuse of the power the law confers.

Q: What did the Aadhaar judgment uphold and strike down?

The majority upheld the Aadhaar Act as a whole, Section 7 conditioning subsidies funded from the Consolidated Fund on authentication, its passage as a Money Bill, Section 59, and Section 139AA of the Income Tax Act on Aadhaar and the Permanent Account Number. It struck down the part of Section 57 enabling bodies corporate and individuals to seek authentication, struck down Section 33(2) on national-security disclosure, read down Section 33(1) to require a hearing, struck down Section 47 in part so an aggrieved individual may complain, curbed transaction metadata, cut authentication-record retention to six months, and quashed the bank-account rule and the mobile-linking directive.

Q: What was the dissent in the Aadhaar judgment?

Justice Chandrachud dissented alone. He held that the Aadhaar Bill did not qualify as a Money Bill because it contained substantial provisions unrelated to the matters enumerated in Article 110, that passing it by that route bypassed the Council of States and was a fraud on the Constitution, and that the Act was unconstitutional in its entirety for that reason alone. He also held the architecture disproportionate: a centralised biometric database, in his assessment, enabled profiling and surveillance, rested on no adequate safeguards, and excluded beneficiaries in its ordinary operation. He directed that the collected data be frozen and, failing fresh legislation within a year, destroyed.

Q: Is the Money Bill question settled?

No. The 2018 majority upheld the certification of the Aadhaar Act as a Money Bill, and that holding stands as the operative law. But a Constitution Bench in Rojer Mathew v. South Indian Bank Ltd. in 2019 held that the majority had not substantially discussed the effect of the word “only” in Article 110(1), and referred the interpretation of the Money Bill definition to a larger bench of seven judges. That reference remains pending and undecided, so the reasoning by which the certification was upheld has been expressly questioned by the Court itself and cannot be described as settled.

Q: How has the privacy judgment been applied since?

It has been applied well beyond identity law. The Constitution Bench in Navtej Singh Johar relied on it to read down Section 377 of the Penal Code, and Joseph Shine relied on it to strike down the adultery offence; the choice of a partner was protected in Shafin Jahan; reproductive autonomy was grounded in it in the medical termination of pregnancy case of 2022; and its proportionality test structured the review of internet restrictions in Anuradha Bhasin, the setting aside of the Reserve Bank’s cryptocurrency circular, and the Court’s response to the Pegasus surveillance allegations. In Kaushal Kishor, the rights under Articles 19 and 21 were held enforceable against non-State actors as well.

Q: Does the privacy right protect against private parties?

The 2017 bench addressed that question directly. The opinions recognised that privacy is invaded by private actors as well as by the State, particularly through the collection and aggregation of personal data, and held that the State bears an obligation to protect informational privacy against non-State actors by putting a data protection regime in place. The right operates against private parties principally through such legislation rather than as a free-standing constitutional claim against them. The later Constitution Bench decision in Kaushal Kishor carried the matter further, holding the rights under Articles 19 and 21 enforceable against persons other than the State.

Q: Who was Justice K.S. Puttaswamy, and why does one name appear on both judgments?

Justice K.S. Puttaswamy is a retired judge of the Karnataka High Court who in 2012 filed the lead petition, Writ Petition (Civil) No. 494 of 2012, challenging the Aadhaar project. Because the challenge turned on whether a fundamental right to privacy existed, his petition generated the reference that the nine-judge bench answered in 2017; when the Aadhaar merits were then decided by five judges in 2018, the case retained his name as its title. Both judgments are therefore cited as Puttaswamy, which is the source of the persistent confusion between the privacy judgment of 2017 and the Aadhaar judgment of 2018.

Q: Why did the privacy question have to go to a bench of nine judges?

Because the authorities blocking it were too large for any smaller bench to move. The Union of India relied on M.P. Sharma, decided by eight judges in 1954, and Kharak Singh, decided by six judges in 1962, each of which had denied that privacy is a fundamental right. Under the Court’s discipline, a decision can be overruled only by a bench larger than the one that delivered it. A bench of nine was therefore constituted in 2017 to reconsider both decisions, and its unanimous answer removed the obstacle and allowed the Aadhaar merits to be heard.

Q: What is the difference between the Puttaswamy judgment of 2017 and the Aadhaar judgment of 2018?

The 2017 judgment, delivered by nine judges and reported at (2017) 10 SCC 1, answered a referred question of principle: it held that privacy is a fundamental right and laid down the test by which invasions of it are judged. It decided nothing about Aadhaar. The 2018 judgment, delivered by five judges and reported at (2019) 1 SCC 1, applied that test to the Aadhaar Act: it upheld the statute’s core, struck down or read down specified provisions and instruments, upheld the Money Bill certification, and produced a lone dissent. The first is a general constitutional decision; the second is its application to one programme.

Q: Who wrote the majority, concurring and dissenting opinions in the 2018 Aadhaar judgment?

The bench comprised Chief Justice Dipak Misra and Justices A.K. Sikri, A.M. Khanwilkar, Dr D.Y. Chandrachud and Ashok Bhushan. Justice Sikri wrote the principal majority opinion for himself, the Chief Justice and Justice Khanwilkar. Justice Bhushan wrote a separate opinion concurring with the outcome, in which he upheld the Act on his own reasoning and differed on some applications. Justice Chandrachud wrote the sole dissent, which would have struck down the Act in its entirety. The result was four to one for the validity of the Act, but the reasoning that binds is the majority’s, and the dissent’s authority is persuasive only.

Q: What did the 2018 judgment decide about metadata and the storage of authentication records?

It cut both back sharply. The majority held that the recording of metadata relating to an authentication transaction, the context of where and for what class of purpose a person authenticated, was impermissible in the form framed by Regulation 26 of the Aadhaar (Authentication) Regulations, 2016, because such records accumulated into exactly the profile the privacy judgment had warned against. It also struck down the permission in Regulation 27(1) to archive authentication records for five years, holding that authentication records are not to be retained beyond six months. The definition provisions of the Act were correspondingly read down to exclude transaction metadata.

Q: Can Aadhaar be demanded for a bank account, a mobile connection or a school place after the judgment?

No. The majority struck down the amendment to the money-laundering rules that had made Aadhaar mandatory for opening and continuing bank accounts, holding it disproportionate. It quashed the Department of Telecommunications directive of 23 March 2017 that had required mobile connections to be re-verified against Aadhaar, because the directive rested on executive instruction rather than law. And it held that Aadhaar cannot be made mandatory for admission of children to school, since elementary education is an entitlement and not a subsidy within Section 7, nor required by bodies conducting examinations such as the Central Board of Secondary Education and the National Eligibility cum Entrance Test.

Q: Did the Aadhaar judgment make Aadhaar voluntary?

No, and the point is often misstated in both directions. The judgment upheld Section 7, under which Aadhaar authentication may be required as a condition of a subsidy, benefit or service funded from the Consolidated Fund, subject to the requirement that a person whose authentication fails be offered alternative means and that no benefit be denied for want of Aadhaar. It also upheld Section 139AA of the Income Tax Act, which makes Aadhaar compulsory for the Permanent Account Number and for returns of income. What the judgment made impermissible was the extension of compulsion beyond that statutory core, to bank accounts, telephones, schools and private transactions.

Q: Where did the proportionality test come from before Puttaswamy adopted it?

The plurality did not invent the structure. A four-pronged proportionality inquiry had been articulated by the Court in Modern Dental College v. State of Madhya Pradesh in 2016, in the context of restrictions on the freedom of trade and profession under Article 19(6): a legitimate purpose, a rational connection between the measure and the purpose, the absence of a less restrictive alternative, and a balance between the measure’s benefit and the right it burdens. Puttaswamy restated that discipline for invasions of privacy under Article 21, in the three requirements of the plurality with Justice Kaul’s safeguards addition, and later benches have stated it in both forms.

Q: Is the right to privacy absolute, or can the State restrict it?

It is not absolute, and the nine judges said so expressly. Privacy, like the other rights of Part III, may be invaded by the State, but only on conditions: the invasion must be authorised by law, must pursue a legitimate State aim, and must be proportionate to that aim, in the sense that the means are rationally connected to the aim and no less intrusive measure would serve, with the procedural safeguards Justice Kaul required. An invasion that fails any of these conditions is unconstitutional. The practical effect is that recognition of the right did not disable government; it fixed the terms on which government must justify itself, and placed the burden of justification on the State.

Q: How many opinions were delivered in the 2017 privacy judgment, and why does that matter?

Six. The plurality opinion was written by Justice Chandrachud for himself, Chief Justice Khehar, Justice Agrawal and Justice Nazeer, and Justices Chelameswar, Bobde, Nariman, Sapre and Kaul each wrote a separate concurring opinion. The result was unanimous, so the holding carries the full authority of the nine-judge bench, but the reasoning is distributed: the test is in the plurality, the safeguards requirement is in Justice Kaul’s opinion, and the fullest treatment of threats from non-State actors is also his. Knowing which opinion carries which proposition matters in practice, because a citation is only as strong as the opinion it is drawn from, and later benches read the judgment accordingly.

Q: What safeguards did the 2018 majority rely on in finding Aadhaar proportionate?

It relied on the limits built into the design as described to the Court and on the Act’s own protections. Enrolment collected a confined set of demographic particulars and core biometrics; authentication returned a yes or no answer without disclosing the purpose of the transaction; the statute prohibited the sharing of core biometric information and criminalised unauthorised access to the repository and unauthorised disclosure of identity information; and Section 7 carried the requirement of alternative identification where authentication failed. Where a safeguard was missing, the majority did not excuse the provision: the national-security disclosure power fell for want of judicial oversight, and the complaint bar fell for denying the individual a remedy.

Q: What happened to the review petitions filed against the Aadhaar judgment?

They were dismissed. In January 2021 a bench of five judges rejected the petitions seeking review of the 2018 judgment by a majority of four to one. The dissenting judge, Justice Chandrachud, who had also dissented in the original case, held that it was a constitutional error to dispose of the reviews while the Money Bill question stood referred to a larger bench, and that the petitions should have awaited the seven judges’ answer. The dismissal closed the review route without closing the underlying question, which continues to rest with the pending reference on the interpretation of Article 110.