Author: Himani Jethwani
College: Jai Narain Vyas University, Jodhpur, Rajasthan
Abstract
Deepfake technology, powered by generative adversarial networks and diffusion models, has moved from a laboratory curiosity to a mainstream threat capable of fabricating hyper-realistic audio-visual content of real individuals. While public discourse on deepfakes has largely centred on privacy invasion, defamation, and electoral misinformation, an equally pressing but underexplored dimension is intellectual property. A synthetically generated video that borrows a celebrity’s face, a musician’s voice, or an artist’s distinctive style raises complex questions of copyright authorship, personality and publicity rights, trademark dilution, and unfair competition. This article examines deepfakes through the intellectual property lens, analysing how existing statutory frameworks in India, the United States, and the European Union grapple with synthetic media, whether copyright law can meaningfully protect a “persona,” and how courts have begun to respond to AI-generated impersonation. It argues that while patchwork protections exist, a coherent, technology-neutral IP framework addressing synthetic likeness is the need of the hour.
To the Point
A deepfake is a piece of synthetic media, video, audio, or image, produced using machine learning techniques to convincingly depict a person doing or saying something they never did. The term is a portmanteau of “deep learning” and “fake.” What began as a niche application on internet forums around 2017 has, within a few years, evolved into an industrial-scale concern touching entertainment, politics, finance, and personal reputation. Deepfakes of film stars endorsing products they never endorsed, synthetic voices of singers performing songs they never recorded, and fabricated speeches of politicians have all surfaced with alarming regularity.
The intellectual property angle is often overshadowed by criminal law and privacy concerns, yet it is arguably where the sharpest legal battles will occur. Three IP regimes are directly implicated.
First, copyright law, because deepfakes are typically built by training algorithms on copyrighted photographs, film clips, or sound recordings, and the output itself may or may not qualify as a copyrightable work with an identifiable “author.” Second, the right of publicity or personality rights, a quasi-property right that allows individuals, particularly celebrities, to control the commercial exploitation of their name, image, voice, and likeness. Third, trademark and unfair competition law, which becomes relevant when a deepfake creates a false impression of endorsement or sponsorship, thereby diluting brand value or misleading consumers.
This article confines itself to these three overlapping strands, while acknowledging that deepfakes also implicate defamation, data protection, and criminal impersonation law, which lie outside the strict IP framework but often travel alongside it in litigation.
Use of Legal Jargon
Understanding the deepfake debate requires familiarity with several IP-specific concepts. “Authorship” under copyright law refers to the person who exercises the requisite skill, judgement, and creative control over a work; this becomes murky when an AI model, rather than a human, generates the final output, raising the “AI authorship problem.””Derivative work,” a term drawn from Section 14 of the Indian Copyright Act, 1957, and Section 106(2) of the US Copyright Act, 1976, denotes a work based upon one or more pre-existing works, relevant because deepfakes are trained on and often reproduce elements of copyrighted source material.
“Moral rights,” recognised under Section 57 of the Indian Copyright Act and Article 6bis of the Berne Convention, protect an author’s right to claim authorship and to object to distortion or mutilation of their work, a right increasingly invoked by actors and singers whose original performances are digitally altered. “Personality rights” or the “right of publicity,” though not codified as a standalone statute in India, has been judicially recognised as an extension of the right to privacy under Article 21 of the Constitution and as a facet of passing off under trademark jurisprudence. In the United States, it is a creature of state common law and statute, most robustly codified in California’s Civil Code Section 3344 and the California Celebrities Rights Act.
“Passing off,” a common law tort actionable under Section 27 of the Trade Marks Act, 1999, in India, protects against misrepresentation that causes damage to goodwill, a doctrine increasingly stretched to cover unauthorised AI-generated endorsements. “Fair use” under Section 107 of the US Copyright Act and “fair dealing” under Section 52 of the Indian Copyright Act provide limited defences for transformative or non-commercial use, defences that deepfake creators frequently, and often unsuccessfully, invoke. Finally, “safe harbour” provisions under Section 79 of India’s Information Technology Act, 2000, and Section 512 of the US Digital Millennium Copyright Act, 1998, shield intermediaries from liability for user-uploaded infringing content, provided they act on takedown notices, a mechanism now central to deepfake removal requests.
The Proof
The statutory and doctrinal foundations for regulating deepfakes through IP law are scattered but discernible.
Under Indian copyright law, Section 13 of the Copyright Act, 1957, protects original literary, dramatic, musical, and artistic works, as well as cinematograph films and sound recordings. A deepfake, by its nature, is derivative: it is trained on and reproduces protected elements of an existing photograph, film, or song. Where the underlying source material is copyrighted and used without licence for training a generative model, the reproduction and adaptation rights under Section 14 are potentially infringed, subject to the fair dealing exceptions in Section 52. Proposed amendments for text and data mining exceptions have not been enacted in India, unlike the European Union, where Articles 3 and 4 of the Digital Single Market Directive (2019/790) carve out text and data mining exceptions that indirectly legitimise certain AI training practices, subject to a rights-holder opt-out under Article 4(3).
Separately, Sections 38 and 38A of the Indian Copyright Act recognise “performer’s rights,” granting performers, actors, singers, and musicians, a right to authorise or prohibit the fixation and reproduction of their performance. A deepfake that synthetically recreates a performer’s voice or visual performance without consent arguably infringes this right, an argument gaining traction as Indian film industries confront AI-cloned voice and likeness misuse.
On personality rights, Indian courts have progressively expanded protection through the Delhi High Court’s interim orders in Titan Industries Ltd. v. Ramkumar Jewellers (2012), which recognised that unauthorised commercial use of a celebrity’s persona amounts to a violation of the celebrity’s right of publicity and constitutes passing off. This line of reasoning was substantially amplified in the 2023-2024 wave of celebrity personality rights suits, including orders protecting the personas of prominent film actors against AI voice-cloning, deepfake merchandising, and unauthorised synthetic content, in which the Delhi High Court granted ex parte injunctions restraining the use of the celebrities’ name, image, voice, and other distinctive attributes, including through AI tools and deepfakes, explicitly citing the risk of morphed and synthetic content.
In the United States, the right of publicity is protected under a patchwork of state statutes. California’s Civil Code Section 3344 prohibits the unauthorised commercial use of a person’s name, voice, signature, photograph, or likeness, while Tennessee’s ELVIS Act of 2024 (Ensuring Likeness, Voice, and Image Security Act) became the first state law explicitly targeting AI voice-cloning, amending the state’s existing personal rights protection statute to expressly cover a person’s voice against unauthorised AI-generated simulations. At the federal level, the proposed NO FAKES Act (Nurture Originals, Foster Art, and Keep Entertainment Safe Act), reintroduced in Congress in recent sessions, seeks to create a federal digital replica right, allowing individuals to control AI-generated replicas of their voice and visual likeness, with statutory damages and safe-harbour carve-outs for platforms that promptly remove infringing content upon notice.
Trademark law supplies a further layer of protection where a deepfake creates a false association with a brand. Section 29 of the Trade Marks Act, 1999, in India, and the Lanham Act, 15 U.S.C. Section 1125(a), in the United States, prohibit use likely to cause confusion as to sponsorship or affiliation. A synthetic video of a celebrity endorsing a product, deployed as an advertisement, would ordinarily fall within these anti-dilution and false-endorsement provisions, as courts have long recognised in analogous “false endorsement” claims involving lookalikes and sound-alikes.
Case Laws
Several judgments, though predating or only recently addressing deepfake technology directly, form the doctrinal scaffolding for future litigation.
Titan Industries Ltd. v. Ramkumar Jewellers, 2012 SCC OnLine Del 2382,
is frequently cited as one of the first Indian decisions to expressly recognise the right of publicity as an enforceable property right, holding that the unauthorised use of a celebrity endorser’s image constituted infringement even absent a direct copyright claim.
ICC Development (International) Ltd. v. Arvee Enterprises, 2003 (26) PTC 245 (Del),
while primarily a trademark ambush-marketing case, laid early groundwork for recognising publicity rights as flowing from the right to privacy.
D.M. Entertainment Pvt. Ltd. v. Baby Gift House, 2010 SCC OnLine Del 4790,
involving the unauthorised sale of dolls resembling a well-known singer, is a foundational precedent establishing that a celebrity’s persona, including caricature and mimicry, merits protection against unjust commercial appropriation.
The 2023-2024 batch of Delhi High Court orders concerning several prominent film actors are the closest judicial engagement with deepfakes specifically. In these matters, the Court restrained defendants from using the actors’ names, images, voices, or other attributes through morphing, GIFs, or artificial intelligence tools, terming such misuse violative of both personality and privacy rights, and in some instances directed the concerned government ministry to identify and block infringing websites.
Internationally, the United States has seen the Ninth Circuit’s decision in Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988), which held that imitating a singer’s distinctive voice for a commercial, even without copying an actual recording, could constitute an actionable misappropriation, a precedent now regularly invoked in the context of AI voice cloning. Similarly, White v. Samsung Electronics America, Inc., 971 F.2d 1395 (9th Cir. 1992), extended the right of publicity to protect a celebrity’s identity even where neither name nor exact likeness was used, involving a robot dressed to evoke a game-show hostess’s persona, a precedent with obvious resonance for AI-generated avatars and synthetic personas.
While no reported Indian or US judgment has yet definitively ruled on a copyright infringement claim arising purely from AI training data ingestion in the deepfake context, ongoing litigation in the United States, including consolidated cases against major AI developers over the use of copyrighted books, images, and music for model training, is expected to shape the eventual contours of liability for the ingestion side of the deepfake pipeline.
Conclusion
Deepfakes sit at an uncomfortable intersection of technology and law, and intellectual property doctrines, though not designed with synthetic media in mind, have proven remarkably adaptable. Copyright law addresses the ingestion of source material and the derivative nature of the output; performer’s rights and personality rights address the unauthorised synthetic replication of a natural person’s voice, face, and mannerisms; and trademark and passing-off principles address the resulting consumer confusion and false endorsement. Indian courts, through an expansive and creative reading of the right of publicity as flowing from Article 21 and from passing off, have moved comparatively quickly to grant injunctive relief to celebrities, a judicial responsiveness that has, for now, outpaced legislative action.
Nevertheless, three gaps remain conspicuous. First, ordinary, non-celebrity individuals, who make up the overwhelming majority of deepfake victims, particularly in cases of non-consensual synthetic intimate imagery, do not enjoy the same commercially-flavoured “persona” protection that courts have extended to film stars, since the right of publicity as currently framed is tethered to commercial goodwill rather than dignitary harm. Second, there is no codified, technology-neutral Indian statute that squarely defines a “digital replica” or synthetic likeness right, leaving courts to stretch existing copyright, trademark, and constitutional doctrine to fit a problem those doctrines were never built to solve. Third, the question of AI training data, whether ingesting copyrighted photographs and films to train a deepfake model itself constitutes infringement independent of the output, remains unresolved in Indian law and is only beginning to be tested in American courts.
In my assessment, the most durable long-term solution lies not in stretching copyright law, which is ill-suited to protecting a “persona” since a person’s face and voice are not themselves “original works of authorship,” but in the enactment of a standalone, statutory digital replica or personality rights law, along the lines of Tennessee’s ELVIS Act or the proposed federal NO FAKES Act, that explicitly covers AI-generated voice and likeness, extends beyond celebrities to all individuals, and is coupled with a robust, deepfake-specific intermediary takedown obligation layered on top of the existing safe harbour framework. Until such legislation is enacted in India, litigants and courts will continue to do commendable but necessarily improvised work, patching together copyright, tort, and constitutional principles to protect what is, at its core, a new kind of property: the right to one’s own digitally reproducible self.
FAQs
Q1. Is a deepfake automatically a copyright infringement?
Not automatically. Infringement depends on whether the deepfake reproduces substantial, copyrighted elements of an underlying photograph, film, or sound recording without licence or a valid fair use or fair dealing defence. A deepfake built from scratch using only a person’s likeness, without copying any specific copyrighted footage, may not infringe copyright at all, even though it could still violate personality or publicity rights.
Q2. Do personality rights exist as a codified law in India?
No. India does not have a standalone personality rights statute. Courts have instead derived the right of publicity from the constitutional right to privacy under Article 21 and from the common law tort of passing off under trademark jurisprudence, as seen in cases such as Titan Industries and the recent celebrity personality rights orders.
Q3. Can an ordinary person, not a celebrity, sue over a deepfake using IP law?
This is legally difficult. The right of publicity, as currently interpreted, is closely tied to commercially exploitable goodwill, which ordinary individuals typically lack. Non-celebrities are usually better served by privacy law, data protection provisions, and criminal remedies for impersonation or non-consensual imagery rather than IP claims.
Q4. What is the ELVIS Act and why is it significant?
The Ensuring Likeness, Voice, and Image Security Act, enacted in Tennessee in 2024, is the first US state law to explicitly extend personal rights protection to a person’s voice against unauthorised AI-generated cloning, directly responding to the misuse of AI voice-cloning technology in the music industry.
Q5. Can platforms be held liable for hosting deepfake content?
Platforms can generally claim safe harbour protection under Section 79 of India’s Information Technology Act, 2000, or Section 512 of the US DMCA, provided they act expeditiously upon receiving a valid takedown notice. Failure to act after actual knowledge of infringing or unlawful synthetic content can result in loss of this immunity.
References (Optional)
- Copyright Act, 1957 (India); Trade Marks Act, 1999 (India); Information Technology Act, 2000 (India).
- Directive (EU) 2019/790 of the European Parliament and of the Council on Copyright and Related Rights in the Digital Single Market (2019).
- World Intellectual Property Organization, WIPO Technology Trends: Generative Artificial Intelligence (2024).
- Titan Industries Ltd. v. Ramkumar Jewellers, 2012 SCC OnLine Del 2382 : (2012) 50 PTC 486 (Del).
- ICC Development (International) Ltd. v. Arvee Enterprises, 2003 (26) PTC 245 (Del).
- D.M. Entertainment Pvt. Ltd. v. Baby Gift House, 2010 SCC OnLine Del 4790.
- Midler v. Ford Motor Co., 849 F.2d 460 (9th Cir. 1988).
- White v. Samsung Electronics America, Inc., 971 F.2d 1395 (9th Cir. 1992).
