Several public figures in India have recently been the subject of deepfakes, some of which were widely circulated on social media. While morphed images are not a new phenomenon, deepfakes created by artificial intelligence (AI) are distinguished by a much higher degree of realism.
As AI—and deepfakes in particular—go mainstream across online platforms, the question of how these technologies affect our legal obligations to one another is gaining salience.
The Delhi high court recently examined the use of non-consensual deepfakes in the matter of Anil Kapoor vs Simply Life India & Others. The Hindi film actor Anil Kapoor had approached the Delhi high court against the misuse of his image, name, voice and persona without his consent.
The Delhi high court observed that tools like AI have made it possible for unauthorized users to use the visual and audio data of any person to create deepfakes. The court held that Anil Kapoor’s likeness, image, persona, etc, all deserve to be protected under the intellectual property law, and therefore users cannot create deepfakes without consent for their own commercial purposes.
The court also observed that the legal protection accorded to free-speech referring to public figures includes satire within its ambit, but does not include speech that jeopardizes the “individual’s personality” or “attributes associated with (them)…”
Television journalist Rajat Sharma has also approached the Delhi high court seeking broad orders on deepfakes, including directions for access to be blocked to software applications that enable their creation. The Delhi high court will consider this case further in July 2024.
Satire versus defamation in the world of deepfakes: While the Delhi high court’s decision in the Anil Kapoor case does provide a Lakshman Rekha, or red line on what is acceptable, the question of how deepfakes interact with the law of defamation remains open.
Defamation, under civil law, has two aspects: first, the publication or appearance of pointed content in any medium, and second, the content that is thus placed in the public domain is found to lower the reputation of its subject.
One line of defence against defamation charges is that it qualifies as ‘fair comment,’ which typically includes forms of satire, parody or mimicry, depending on the specific statements made and context in which the content is published. The same standards cannot be directly applied to deepfakes.
In the case of satirical drawings and depictions in film or text of an individual, a fair-minded viewer or reader is aware that the content in question did not actually originate from the subject of the satire. However, recent research suggests even fair-minded people may have difficulty determining whether statements or actions are attributable to the subject of a deepfake.
However, if a deepfake creates the illusion that its subject has said or done something that would harm the person’s reputation if publicized, it would arguably no longer qualify as an opinion or satire. Defamation may arise not only from the content of such a deepfake, but also from the mis-attribution of words and actions to the person depicted.
Celebrities famous for promoting a healthy lifestyle or green causes could have their reputation harmed by deepfakes that depict them discussing a love for junk-food or flying on a private jet. These situations appear to fall outside the ambit of protected satire, recognized by the Delhi court in the Anil Kapoor case.
Balanced use of deepfakes for the creation of satire: While several jurisdictions, including India, see a legitimate public interest in satire, the obvious signs that usually leads an audience to conclude that the content they’re consuming is satirical are likely to be missing in hyper-realistic deepfakes. The use of disclaimers on deepfakes can mitigate the risk that a viewer will be misled into taking these as true camera-filmed depictions.
The EU now requires AI-generated content, including deepfakes, to be labelled as such. A similar measure is included in advisories issued by India’s ministry of electronics and information technology on 1 March and 15 March 2024.
Indian courts have held that even where content does not, from its context, appear to be defamatory, a disclaimer indicating that the content is satire may usefully safeguard the rights of all parties.
As deepfakes become increasingly likely to pass for genuine portrayals, content that is meant to parody rather than deceive may still be defamatory. Clear warnings that such satirical content is generated using AI, and that it does not actually seek to depict real events, could help manage viewer expectations and reduce the likelihood of unintended defamation charges and their penal consequences.
Hardik Malik, associate at Khaitan & Co, contributed to this article.