BORROWED FACES, STOLEN VOICES: INDIA’S LAW AND THE DEEPFAKE AGE

Author: Mahi Sinha, Lloyd School of Law 

 

An inquiry into whether a legal order built for flesh, ink and signature can govern the counterfeit self

 

PREFATORY REMARKS: THE FORGED PERSONA 

It is a premise so elementary that the law seldom troubles to state it: that every individual is the sole author of their own appearance. The countenance, the timbre of the voice and the cadence of speech are the instruments by which a person is identified, credited and trusted in the society. Historically, the law had little occasion to test this premise, for the simulation of a human being lay beyond the competence of the ordinary wrongdoer. A forger might counterfeit a signature; he can’t counterfeit a person. 

 

Synthetic media has removed that immunity. By means of generative adversarial networks, diffusion models and voice cloning software, it is now possible to attribute to any identifiable individual words never uttered and acts never performed, with verisimilitude that eludes the lay observer and, with growing frequency, the expert as well. The mischief is not the preserve of the celebrated. It descends most severely upon the obscure: the student whose photograph is transposed onto obscene imagery, the elderly parent deceived by cloned voice soliciting money, the professional woman whose standing is held to ransom by fabricated recording. 

 

The question for Indian jurisprudence is accordingly doctrinal before it is technological does the existing legal order possess the conceptual vocabulary to identify what has been taken from the victim? This article submits that the answer is a qualified affirmative. The materials for a remedy are scattered through the constitution, the penal statutes, the law of information technology and the common law of torts, yet they lack a unifying principle. That principle, it is argued, is consent. 

 

CONSENT AS INCIDENT OF CONSTITUTIONAL AUTONOMY 

In private law, consent is a modest concept, the hinge of the contract and the shield of the surgeon. In constitutional law it assumes a grave office. It is the juridical expression of self sovereignty: to consent is to be demoted from a subject of rights to an object of use. 

Indian constitutional doctrine reached this understanding by degrees. In R.Rajgopal v. State of Tamil Nadu(1994), the supreme court held the right to privacy to be implicit in the guarantee of life and personal liberty under article 21, and affirmed that a citizen may protect the privacy of the person, family and home against publication without consent, save in matters of public record. The nucleus of the consent principle was thus already present: the publication of the self required the licence of the self. 

 

The doctrine attained maturity in Justice K.S Puttaswamy (retd.) v. Union of india (2017), where a bench of nine judges declared privacy to be a fundamental right rooted in dignity, liberty and autonomy. Two strands of that decision bear directly upon synthetic media. The first is informational self-determination, namely the claim of the individual to govern the circulation of information concerning themselves. The second is the recognition that privacy shelters the choices through which a person fashions their identity. A deepfake offends both, for it conscripts the self as raw material and then publishes a fabricated narrative under the self’s name. 

The later authorities on personal liberty and bodily integrity fortify the point. Autonomy, in this idiom, is not a privilege of the fortunate but the very precondition of dignity. A person whose likeness may be pressed into any performance, however degrading, without their knowledge has been deprived of dignity in its most rudimentary form. 

 

The constitutional premise is therefore secure. What remains in doubt is its range of operation. Fundamental rights lie, in the first instance, against the state, and their horizontal application against private wrongdoers, though judicially acknowledged in particular settings, requires statutory and common-law machinery to acquire practical effect. It is to that machinery that the inquiry must now turn. 

 

THE STATUTORY MOSAIC

India has enacted no provision that designates the deepfake as a distinct offence. What it possesses is a mosaic of general provisions, each of which captures a fragment of the injury and none of which captures the whole. 

 

The information technology act 2000, section 66C penalises the fraudulent or dishonest use of the electric signature, password or other unique identification feature of another, while section 66D penalises cheating by personation by means of a computer resource. Both are plausibly attracted where a cloned voice or face is deployed to defraud. Section 66E punishes the capture, publication or transmission of the image of a private are of any person without consent, and section 67 and 67A proscribe the publication or transmission of obscene and sexually explicit material in electronic form. These provisions have served as the principal instruments of cyber prosecution and are not without utility against sexualised forgeries. 

 

Their deficiency, is however, structural. Section 66E presupposes the capture of a genuine image; a synthetic image captures nothing, because no private area was ever exposed to the lens. Sections 67 and 67A, for their part, are directed at the obscenity of the material, not at the violation of the person portrayed, so that the gravemen of the offence is public morality rather than the victim’s want of consent. The statute thus interrogates the wrong question. It asks whether the content is lewd, when the injury resides in the circumstances that the content is false and unconsented. 

 

The Bharatiya Nyaya Sanhita, 2023. The successor penal code furnishes further, though imperfect, resources. The provisions relating to defamation, cheating, cheating by personation, forgery undertaken for the purpose of harming reputation, criminal intimidation, stalking and the voyeurism of women may each be invoked as the facts permit. A deepfake employed to extort is extortion in substance; one employed to impersonate a corporate officer so as to procure a remittance is cheating by personation. Yet the pattern recurs: the law reaches the consequence of the counterfeit and not the counterfeit itself. Where a synthetic image occasions humiliation without demonstrable fraud, intimidation or defamation, the penal code is apt to fall silent.

The Copyright Act, 1957. A less noticed avenue lies in the law of performers’ rights. Section 38 vests exclusive rights in the performer, and Section 38B secures to the performer the right to claim credit and to restrain any distortion, mutilation or modification prejudicial to their reputation. This is a serviceable instrument for actors, vocalists and public performers whose work or persona is manipulated. But it presupposes a performance and a performer; the schoolgirl and the clerk have no performance to protect.

The Digital Personal Data Protection Act, 2023. Here, for the first time, consent appears as an express statutory category. The Act conditions the lawful processing of personal data upon the free, specific, informed, unconditional and unambiguous consent of the data principal, subject to enumerated legitimate uses. Facial and vocal data processed to train or operate a generative model is, prima facie, personal data, and the scheme is the nearest approach the statute book makes to a codification of informational self-determination.

The Act, however, carries exemptions of considerable breadth, including for instrumentalities of the State and, more pertinently, for personal data processed by an individual for a personal or domestic purpose. The private person who fabricates a humiliating video of a classmate or colleague may well shelter behind a carve-out never designed for his benefit. Furthermore, the Act regulates the processing of data by data fiduciaries; it is a statute of compliance and not a charter of personal remedy against a malicious stranger.

JUDICIAL INNOVATION: THE DOCTRINE OF PERSONALITY RIGHTS

Where the legislature has hesitated, the high courts have been inventive. From titan industries ltd. v. Ramkumar jewellers (2012) onwards, and through a succession of orders of the Delhi and Bombay high courts, Indian courts have fashioned a body of personality rights, anchored in the torts of passing off, in the right to privacy and in the proposition that a person’s name, image, voice and distinctive attributes constitute an interest which that person alone may exploit. 

in Anil Kapoor v. Simply Life India (2023), the Delhi High Court restrained the unauthorised use of the actor’s name, likeness, voice and mannerisms, and took express note of the perils of artificial intelligence and deepfake technology. In Arijit Singh v. Codible Ventures LLP (2024), the Bombay High Court granted relief against the use of AI voice-cloning tools to reproduce the vocalist’s voice. Orders in favour of other public figures have since followed, several of them directed not merely at the named defendants but also at intermediaries, with a mandate for the removal of the offending content.

These pronouncements merit sincere admiration. They demonstrate the capacity of the common law to meet novelty without awaiting the slower tempo of legislation, and they confirm that the voice and likeness of a person are, in the eyes of the Court, subjects of legal protection.

Yet the jurisprudence bears a conspicuous bias: it is, in the main, a jurisprudence of the eminent. Its reasoning is frequently commercial. The celebrity holds goodwill; the deepfake dilutes or misappropriates it. That is the logic of passing off, which presupposes a goodwill to be injured and a public liable to be deceived. For the ordinary citizen, whose features command no market but whose dignity is no less precious, the reasoning fits awkwardly. The law protects the monetised self more readily than the merely human one.

The resulting paradox is plain. The litigant best endowed with the means to seek relief enjoys the most developed remedy, while the litigant of modest means is relegated to the blunt instruments of the criminal process.

THE INTERMEDIARY AMMENDMENT OF 2026

The Executive has not been idle. Following earlier advisories to platforms, the Ministry of Electronics and Information Technology notified, on 10 February 2026, amendments to the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, bringing what it terms synthetically generated information within the due diligence framework of the Rules.

The definition is technological and perceptual in character: audio, visual or audio-visual material artificially or algorithmically created or altered so as to appear authentic and to portray persons or events as indistinguishable from the genuine. The amendment mandates prominent labelling and provenance metadata, imposes upon significant social media intermediaries obligations of user declaration and technical verification, and, most conspicuously, abbreviates the timelines for takedown. As reported in professional commentary, the general period for acting upon unlawful content has been reduced from thirty-six hours to three, with shorter periods still for non-consensual intimate imagery.

As a matter of institutional design the measure is a genuine advance, and three observations suffice.

First, it relocates the point of regulation from the wrongdoer, who is commonly anonymous and untraceable, to the platform, which is identifiable and answerable. The choice is pragmatic, for a law that can reach only the invisible seldom reaches anyone.

Second, it preserves the architecture of safe harbour under Section 79 of the Information Technology Act, conditioning immunity upon due diligence. The intermediary is not made the insurer of all content; it is made accountable for the quality of its response.

Third, and most material to the present inquiry, the amendment is notable for what it omits. The Rules regulate labelling and removal. They confer no right upon the person depicted. A label informs the viewer that an image is synthetic; it does not inform the law that the image was fabricated without the consent of the person it portrays. A scrupulously labelled forgery of a private citizen remains an affront to that citizen, notwithstanding its candour as to its own artificiality. The regime is one of transparency, whereas the injury is one of appropriation.

There are, besides, the familiar misgivings that attend any compressed takedown regime. A three-hour window invites over-removal by risk-averse platforms, and Shreya Singhal v. Union of India (2015) reminds us that intermediary liability must be reconciled with Article 19(1)(a). The technology that wounds also sustains satire, parody and political commentary, forms of expression which a constitutional democracy must tolerate. A rule incapable of distinguishing a malicious forgery from a caricature of a minister will, in due course, be invoked to suppress the latter.

THE RESIDUAL LACUNAE 

Having surveyed the field, one may identify with some precision the points at which the present law fails.

Absence of a substantive right in the person depicted. No Indian statute declares, in terms, that an individual holds a right against the creation and dissemination of realistic synthetic depictions of themselves without consent. Such rights as exist are derivative: of privacy, of passing off, of the law of obscenity, of data protection. Derivative rights are brittle, since they depend upon the plaintiff’s ability to fit a novel wrong into an older mould.

The problem of mere creation. Most existing provisions bite upon publication or transmission. A deepfake that is made and retained, but not yet circulated, occupies an uncertain position, although the injury to dignity may already have been complete and the peril of dissemination is ever present. A law of consent, rightly understood, would condemn the unauthorised making no less than the unauthorised showing.

The taxonomy of harm. Because the penal provisions cluster around obscenity, fraud and defamation, deepfakes that are neither sexual, nor fraudulent, nor defamatory, such as a fabricated endorsement or a synthetic political statement, fall into interstitial space. The law lacks a category for the simple wrong of being made to say what one never said.

The anonymity of the offender. Relief against a faceless account is of limited practical value. The identification of offenders depends on platform cooperation, cross-border requests for data and forensic capacity, none of which is abundant in Indian enforcement.

The inequality of access. Finally, there is the asymmetry already noted: the dignity of the affluent is protected by swift injunctive relief, that of the ordinary citizen by a slow and often forbidding criminal process.

TOWARDS A JURISPRUDENCE OF DIGITAL PERSONHOOD

What, then, is to be done? The answer lies neither in hasty prohibition, which would imperil legitimate expression, nor in patient reliance upon the existing mosaic, whose seams have already shown. It lies in the articulation of a coherent principle, and it is respectfully submitted that the principle is consent as the measure of digital personhood.

The proposition may be stated shortly. The likeness, voice and characteristic manner of a person are extensions of the self. The creation or circulation of a realistic synthetic depiction of an identifiable person, without that person’s free and informed consent, violates their autonomy and dignity, irrespective of whether the depiction is obscene, fraudulent, defamatory or commercially injurious. Such a principle would not displace the existing offences; it would underlie them as their common rationale.

Several consequences follow.

First, the principle would sustain the recognition, whether by statute or by judicial development, of a general personality right unconditioned by fame or goodwill. The Courts have already erected the scaffolding; what is required is to detach it from the commercial logic of passing off and to re-found it upon Article 21.

Second, it would settle the standard of consent. Borrowing from the data protection statute, consent ought to be free, specific, informed and revocable. Consent to be photographed is not consent to be synthesised, and consent to one synthetic use is not consent to every conceivable use. A person who licenses their voice for an advertisement has not thereby licensed it for a political broadcast. The law of contract, with its settled wariness of unconscionable and over-broad bargains, supplies useful analogies, particularly in view of the growing practice of requiring performers to surrender digital replica rights in standard-form agreements.

Third, it would permit a calibrated treatment of exceptions. Consent is not absolute; it must yield to the legitimate claims of speech. Parody, satire, criticism, news reporting and bona fide artistic expression ought to be protected, subject to clear disclosure and to the proportionality analysis which the Supreme Court has long demanded of restrictions on Article 19(1)(a). The test in Puttaswamy, requiring legality, a legitimate aim and proportionality in any limitation of a fundamental right, supplies the appropriate template. The balance must be struck by principle, not by the tempo of a takedown clock.

Fourth, it would demand a distinct regime for the gravest category: non-consensual sexual deepfakes. Here no balancing is called for, and the law should say so without equivocation. The harm is grievous, the expressive value negligible, and the victims overwhelmingly women and, increasingly, children. Specific criminalisation, victim-centred procedure, expeditious injunctive relief and protection of anonymity during proceedings are all warranted. The solicitude shown in Subhranshu Rout v. State of Odisha (2020), where the Orissa High Court acknowledged the victim’s interest in the removal of such material from the internet, deserves to be generalised and not treated as an exceptional indulgence.

Fifth, it would encourage provenance by design. The labelling regime of 2026 is a commencement, but the more durable solution lies in technical standards of authentication permitting genuine media to establish their origin, so that the burden shifts by degrees from debunking the false to certifying the true.

None of this need await a comprehensive new enactment, although the long-anticipated successor to the Information Technology Act might, if carefully drafted, furnish its natural home. In the interim the Courts may begin the work. The constitutional text already contains the necessary premises, and the common law has never required legislative leave to recognise a wrong.

CONCLUSION: THE DIGNITY OF UNREPRODUCED 

There is an ancient maxim that where there is a right, there is a remedy. The deepfake obliges us to invert it: where there is a wrong, we must discover the right which it violates. The wrong is plain enough. A person has been made to appear, to speak and to act in a world they never entered, and the law, for want of an apt term, has hesitated to say that they have been wronged.

Indian law is not wholly unequipped for the encounter. It possesses a constitutional doctrine of dignity and informational autonomy; a body of personality-rights decisions which has anticipated the problem; a penal code and cyber statute that reach its worst consequences; a data protection framework that speaks, at last, the language of consent; and an intermediary regime that has begun to exact accountability from the platforms which amplify the harm. But these are instruments without a score. They await a conductor.

That conductor is the principle of consent, understood not as a technicality of contract but as the legal form of human sovereignty over the self. If Indian jurisprudence will consent to be guided by it, the existing law, supplemented by modest and considered reform, may yet prove equal to the protection of digital personhood. If it will not, we shall find ourselves in the paradoxical position of possessing a Constitution that guarantees dignity and a legal order that cannot recognise its theft.

FAQS:

1. Is there a dedicated statute against deepfakes in India?

No. India has no enactment that names the deepfake as a distinct offence. Remedies must be assembled from the Information Technology Act, 2000, the Bharatiya Nyaya Sanhita, 2023, the Copyright Act, 1957, the Digital Personal Data Protection Act, 2023, and the constitutional right to privacy under Article 21.

2. Which provisions are most commonly invoked against a deepfake?

Sections 66C and 66D of the Information Technology Act (identity theft and cheating by personation), Sections 66E, 67 and 67A (violation of privacy and obscene or sexually explicit material), and the Sanhita’s provisions on defamation, cheating, forgery for harming reputation, criminal intimidation and stalking. The applicable provision depends on the use to which the deepfake is put.

3. Can an ordinary person claim personality rights, or are they confined to celebrities?

The reported orders have largely favoured public figures, because the reasoning draws on goodwill and passing off. An ordinary citizen may nonetheless invoke the right to privacy and dignity under Article 21, the law of defamation, and the general jurisdiction of civil courts to grant an injunction. The position for non-celebrities is less settled, which is precisely the lacuna this article examines.

4. What does the 2026 amendment to the Intermediary Rules require?

As reported in professional commentary, it brings synthetically generated information within the due diligence obligations of intermediaries: prominent labelling, provenance metadata, user declarations and technical verification by significant social media intermediaries, and markedly shorter takedown periods. It does not, however, create a substantive right in the person depicted.

5. Does the Digital Personal Data Protection Act, 2023 protect against deepfakes?

Partly. Facial and vocal data are personal data, and the Act is founded on consent. But it regulates data fiduciaries, contains broad exemptions, and excludes processing by an individual for a personal or domestic purpose, so it is an imperfect remedy against a private individual who fabricates a deepfake.