Back to the eighties… who owns your AI-generated retro self?

Social media has discovered a new time machine. Over the past few days, Instagram and X have been flooded with photographs of people apparently transported back to the 1980s. A contemporary selfie is uploaded to an AI tool and seconds later, it returns with voluminous hair, aviator glasses, vintage clothing, warm studio lighting, and the unmistakable grain of an old photograph.

The trend has travelled remarkably quickly. Celebrities, influencers, politicians, and ordinary users have joined in. Some have even responded by posting genuine photographs from the 1980s. What appears to be harmless digital nostalgia, however, presents an interesting legal question: who owns your AI-generated 1980s self?

The answer is more complicated than it first appears. Behind one viral image may lie at least three distinct subjects of legal protection: the photograph uploaded by the user; the AI-generated image produced from it; and, potentially, pre-existing photographs, films, costumes, characters, trademarks, logos, or other copyrighted works reflected in the output.

India’s Copyright Act, 1957 was obviously not drafted with generative AI in mind. Yet one provision inserted decades before today’s generative AI boom makes the Indian position particularly interesting.

The first misconception worth addressing is that uploading a photograph to an AI tool somehow removes copyright from the equation. It does not. A photograph is expressly an “artistic work” under Section 2(c) of the Copyright Act. Section 2(d)(iv) ordinarily identifies the person taking the photograph as its author, while Section 17 determines who will be the first owner subject to its statutory exceptions.

The person appearing in a photograph and the person owning the copyright of it are therefore not necessarily the same.

A selfie taken by the user presents a relatively straightforward situation. But suppose a professional photographer took the original photograph. Or the image was taken during a commercial photoshoot. Or, more obviously, someone downloads a celebrity photograph from the internet and asks an AI tool to transform it into an imaginary 1980s portrait. The fact that AI subsequently modifies the photograph does not make the underlying copyright disappear.

Section 14 gives the copyright owner exclusive rights that include reproduction and adaptation of an artistic work. Section 51 deals with unauthorised acts falling within those exclusive rights. The first question in any AI-image analysis should therefore sometimes be not “who owns the AI image?” but “did the user have the right to use the input image in the first place?”

This is where Indian law becomes particularly interesting. Section 2(d)(vi) of the Copyright Act provides that for a literary, dramatic, musical, or artistic work that is “computer-generated”, the author is “the person who causes the work to be created”. That language was introduced long before modern generative AI. Yet it potentially gives Indian law a statutory starting point for a problem with which jurisdictions around the world are now struggling. The provision has also just received significant attention from the Indian Copyright Office.

In its August 2026 decision concerning Stephen Thaler’s AI system DABUS and the artwork A Recent Entrance to Paradise, the Copyright Office drew an important distinction between creation and legal authorship. On the facts before it, the Copyright Office found that the particular AI-generated work satisfied the originality requirement. However, it refused to recognise DABUS itself as the author. An AI system is not the “person” contemplated by Section 2(d)(vi). Instead, the inquiry is directed towards identifying the legally recognised person who caused the particular work to be created.

Importantly, the Copyright Office did not say that every person who types a prompt automatically becomes an author. The Copyright Office instead examined the degree of human involvement behind the particular output, including who provided the inputs, controlled the process, and ultimately set the creation in motion. That distinction could become crucial for viral AI image trends.

Consider two users.

User A uploads a selfie and types: “Make this look 1980s.” The AI determines virtually everything else, such as the hairstyle, clothing, lighting, background, pose, colour palette, and composition.

User B takes a very different approach. They select the source photographs; write detailed prompts specifying composition, clothing, lighting, and background; reject multiple generations; modify the instructions repeatedly; select particular elements; and continue iterating until a specific final image is produced.

Are both users equally the “person who causes the work to be created”? The statutory language does not provide a mechanical answer. The Supreme Court’s decision in Eastern Book Company v D.B. Modak (2007) remains relevant. Indian copyright law does not demand novelty, but originality requires independent creation accompanied by at least a minimal degree of creativity. Purely mechanical labour is insufficient.

The recent DABUS decision suggests that the absence of human control over every pixel does not necessarily defeat originality. But the harder cases concern the opposite end of the spectrum: where human involvement consists of little more than a generic instruction followed by acceptance of whatever the machine produces.

In such cases, there may be a real question whether the user exercised sufficient creative control to claim authorship or whether the output sits uncomfortably between technological creation and the Copyright Act’s human-centred concept of authorship. That question will eventually require judicial clarification.

There is, however, another important distinction. Nobody owns the 1980s aesthetic. Copyright protects particular expression, not a general idea, theme, period, or aesthetic. Big hair, warm photographic tones, colourful clothing, film grain, aviator sunglasses, and a generic ‘1980s Bollywood’ atmosphere do not become the monopoly of one photographer, filmmaker, or studio. This distinction is important because generative AI debates can sometimes collapse style and copyright into the same concept.

Asking AI to make a photograph “look like an Indian film portrait from the 1980s” is therefore fundamentally different from asking it to reproduce a particular photograph, film poster, character, or identifiable creative composition. The legal risk rises as the prompt moves from era to expression.

“Give me an 80s Bollywood look” is one thing. “Make me look exactly like the protagonist in this particular film poster, wearing the same costume, standing in the same pose against the same background” is quite another.

The closer an output comes to reproducing protectable expression from an existing work, the stronger the potential infringement argument becomes.

There is another popular assumption worth testing: if AI changes an existing image substantially, the result must be lawful because it is ‘transformative’. India’s Copyright Law does not contain a general, free-standing transformative-use defence. Section 52 provides specific exceptions, including fair dealing for private or personal use, research, criticism, or review, and reporting current events and current affairs.

Consequently, a private experiment with an image may present a very different risk profile from a commercial social media campaign. Imagine that a user creates a humorous AI portrait resembling a famous 1980s movie poster and keeps it on their phone. Now imagine that a fashion company produces the same image featuring its products, places its logo on it, and distributes it through paid Instagram advertising. The technological act may be similar. The legal context is not.

The purpose, source material, extent of copying, and manner in which the resulting work is exploited can materially affect the analysis.

The trend also exposes an important distinction between copyright and personality rights. A person does not necessarily own copyright merely because their face appears in a photograph. However, the unauthorised commercial exploitation of someone’s identity may engage a separate body of rights.

Indian courts have increasingly protected commercially valuable aspects of personality, name, image, likeness, voice, and other distinctive attributes against unauthorised exploitation. In several cases, the Delhi High Court has dealt expressly with morphed images, AI-generated material, GIFs, and other technological uses of the actor’s persona. It recognised that modern technological tools make it extraordinarily easy to imitate a celebrity’s identity and restrained unauthorised exploitation of a celebrity’s name, likeness, image, voice, and personality, particularly in commercial contexts. Recent personality rights cases have continued that trajectory.

Uploading your own photograph and transforming yourself into an imaginary 1980s film star is therefore one proposition. Uploading a celebrity’s photograph, generating a fictional endorsement, and using the resulting image to advertise a business is an altogether different proposition. The latter may raise personality and publicity rights, passing off, privacy, and possibly trademark and copyright issues simultaneously.

The AI may also introduce something the user never expressly requested. Imagine an eighties portrait in which the AI adds a recognisable luxury handbag, sportswear logo, automobile badge, or distinctive product packaging. Copyright may not be the only intellectual property (IP) regime engaged. Where a registered trademark is reproduced in a manner suggesting a commercial association, sponsorship, or endorsement, trademark infringement and passing-off questions can arise. The risk becomes considerably more significant where an influencer or business uses an AI-generated image commercially. Generative AI therefore produces an unusual possibility: the user may inadvertently publish an IP reference that was generated rather than deliberately selected.

That is a practical reason why businesses should subject AI-generated advertising material to human IP clearance before publication.

Copyright is also no longer the only statutory framework relevant to such images. India’s Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021 were amended in February 2026 specifically to address “synthetically generated information” (SGI). The amended framework introduces obligations relating to synthetic content, including the labelling and metadata or identifiers in relevant circumstances, together with enhanced intermediary due diligence requirements. Significant social media intermediaries face additional obligations concerning declarations, verification, and prominent labelling of SGI.

This represents an important regulatory shift. The legal response to AI imagery is moving beyond the traditional question of “who owns this image?” towards another question...

For a fun retro portrait, this may not matter much. However, for a fake celebrity endorsement, political image, advertisement, or deepfake, it can have serious legal consequences. Privacy and data protection considerations may also arise when identifiable photographs are uploaded to AI systems, particularly where photographs contain personal data relating to persons other than the user. India’s digital personal data protection framework adds another developing layer to the analysis, although its substantive provisions are being brought into force in stages.

Indian law does not require social media users to abandon the fun of AI nostalgia. It does, however, suggest some sensible boundaries. Users should preferably upload photographs they have taken themselves or are authorised to use. Prompts should seek an era or general aesthetic, rather than reproduction of a particular copyrighted photograph, poster, or artwork. Celebrity likenesses should not be used to imply endorsements without consent.

For businesses, the standard should be higher. AI-generated content intended for advertising should be reviewed for third-party photographs, copyrighted characters and artworks, trademarks, celebrity likenesses, and misleading synthetic content before publication. Businesses should also preserve the prompts, iterations, and source material used to produce important commercial assets. Those records may eventually prove valuable in demonstrating how the work was created and who ‘caused’ it to be created.

And where commercially significant AI-generated artwork is being commissioned, contracts should expressly deal with ownership, permitted inputs, third-party rights, and responsibility for infringement rather than assuming conventional copyright standards will answer an unconventional technological problem.

At first sight, the viral eighties portrait seems to be one of social media’s more harmless diversions. Legally, however, it captures almost every unresolved tension in generative AI. The input may already belong to somebody. The output may or may not qualify for independent copyright. The person entering the prompt may or may not be the statutory author. The aesthetic may be free for everyone to imitate while particular expression within that aesthetic remains protected. A person’s face may engage personality rights even though that person does not own the photograph. And an innocent-looking AI image can contain trademarks, copyrighted material, or synthetic representations that create an entirely different set of legal consequences when used commercially.

India is unusually well placed to confront at least part of this puzzle. Section 2(d)(vi) of the Copyright Act anticipated “computer-generated” works decades before today’s generative AI systems existed and the Copyright Office’s recent DABUS ruling has now begun the process of applying that language to modern AI. However, the viral trend also demonstrates the limit of asking simply whether AI can own copyright.