A question of integrity in the time of AI

'Don’t give away the ending. It is the only one we have.’ With these words, Alfred Hitchcock devised a canny marketing strategy for his landmark 1960 film Psycho even as he created possibly the world’s first spoiler alert warning. Cut to 65 years later and we find ourselves in an era of AI-generated deepfakes, where it turns out that filmmakers and viewers no longer worry about a particular end to a film being the only one they have.

Twelve years after its original release, Aanand L. Rai’s tragic romance Raanjhanaa was re-released this year with a different ending, one generated entirely by AI. If in the original, the protagonist Kundan (played by Dhanush) is shown being shot and dying of his injuries, in the revised AI version, we are given a happy ending in which Kundan survives and wakes up in a hospital bed.

If it were a product of AI hallucination, this alternative ending might have been amusing, but instead it feels a little sinister, especially since both the director and star objected to it and are considering legal action against the producer for creating this version without their consent.

There are two dimensions to their objections that come into play. The Copyright Act recognises a special class of rights (moral rights) that survive even the transfer of ownership of rights. These include the relatively straightforward ‘right of paternity’ or the right to be identified as the author of a work (a publisher may own the copyright but cannot name someone else as the author), and the far more ambiguous ‘right to integrity’ of a work.

The right to integrity protects the moral coherence of a work from acts that distort, mutilate or modify the work in a manner that prejudices the honour or reputation of the author. Thus, even if you are lucky enough to be the owner of a Husain painting, you would not be entitled to destroy or mutilate the painting without violating the painter’s moral rights over his creation. On that count, it would appear, Aanand L. Rai may have a straightforward claim against the AI-generated ending of Raanjhanaa.

What about Dhanush? He was an actor in the film, but owns no copyright over it. While there is a special class of rights accorded to performers, the question of whether performers are entitled to claiming moral rights over their performance is far more ambivalent.

In the past few years, we have seen a spate of cases with celebrities approaching courts to protect their ‘right to personality’ against misappro-priation by AI companies (who use deepfakes to produce an unauthorised doppelganger of the star or use large language models to learn the voice of a singer and replicate the singer’s voice and style). Thus, celebrities from Amitabh Bachchan to his son Abhishek and daughter-in-law Aishwarya to Rajinikant to Arijit Singh have managed to obtain relief from courts against AI-generated impersonations of them.

The ‘right to personality’ is a conceptually intriguing right as it brings together diverse philosophical traditions within intellectual property theory — traditions that are often at loggerheads with each other.

There are broadly two justificatory theories within copyright. The Anglo-Saxon model, best represented by John Locke, emphasises the economic argument of incentives, arguing that in the absence of exclusive property rights in the intangibles, there would be no incentive to create anything new at all. This theory is primarily economic in nature and focuses on IP as a property right.

The continental tradition, on the other hand, draws from Immanuel Kant to emphasise the importance of the intangible personality of a creator and justifies copyright as a system that protects the autonomy of the creator’s identity, and sees IP as an author’s right.

The right to personality rests somewhat uncomfortably between these two theories. It is derived primarily from the Kantian tradition of personal autonomy and is an extension of the right to privacy. On the one hand, it is technically available to every individual and meant to protect the integrity of an individual’s identity. At the same time, as a legal right, it can only be exercised by someone whose identity has a commercial value — such as a celebrity.

The right has primarily been driven by the global rise of celebrity culture, widespread media influence and the proliferation of advanced digital exploitation technologies.

The normative justification of the right is based on several paradoxes — an underlying idea that one’s identity is an inalienable possession even as its legal status rests on the ability to commodify and alienate one’s identity. It flows from the right to privacy, even as it is commercially derived from the public status of a celebrity.

How do we read conflicts over the right to personality in the era of AI? If one were to move beyond the obvious question of the economic interests of celebrities and ask instead if there is a perspective on this right from the vantage of the ordinary viewer or spectator, what would these conflicts illuminate for us?

There is one overwhelming question that cuts though all debates over AI and that is the question of what it means to be human in a time when machines have replicated many functions that we as humans had conceitedly described as being unique to human beings — the capacity for language, intelligent thought and creativity.

A variant of this question echoes in the perplexing experience of our capacity to be moved by a fictional performance, as seen in cinema. This is an age-old question in philosophy and aesthetics that illuminates the profound power of art to create an emotional reality that can feel more intense than actual events.

This affective experience hinges on a ‘willing suspension of disbelief’, which in turn is produced by the authenticity of a performance. A truly great performance is one where we react emotionally despite our awareness of it being a performance. Think of Rajesh Khanna’s death in Anand, for instance, or Dumbledore’s in the Harry Potter series.

Readers and spectators have always claimed agency over the meaning of a work through active appropriation such as fan fiction, but the discomfort we feel with AI-generated alternative endings seems to be fundamentally different from the tradition of fan fiction, even though both involve manipulating existing material.

Is our concern simply about manipulation of the original experience? The existence of a director’s cut for iconic films such as Blade Runner or Sholay seems to belie this claim. But online, you find innumerable instances of digital artefacts — occasionally funny and often troubling versions of AI-generated sequences of iconic films (a Thakur with hands intact etc.)

During the Cold War, it was said of the censorship regimes in Russia and the USA, that nothing goes in Russia because everything matters, while everything goes in America because nothing matters. It seems our discomfort stems not from alternative versions, but from something about the mechanism of creation. AI’s lack of human intention, empathy and moral consideration seems to trouble us.

We can argue that it matters to Aditya Chopra that Amrish Puri would let go of Kajol’s hands in the climax of DDLJ, allowing her to live her life with freedom; it matters to us that Jai sacrifices his life for Veeru, but does it really matter to AI and might that be the core question about what the integrity of a work means in today’s time?

Lawrence Liang is a lawyer and academic who has worked extensively on intellectual property rights