Behind the Pixels: Exploring the Technology of Synthetic Media and India's Legal Battle with Deepfakes

Introduction

A short time ago, popular rapper Eminem left the audience in awe when he performed as his alter ego, Slim Shady at the MTV’S Video Music Awards .This performance, while exceptional, represents just one side of the deepfake technology, which is not as astonishing as it seems but has been weaponized to cause harm. In India, celebrities like Rashmika Mandanna, Alia Bhatt, and Katrina Kaif have faced deepfake-based controversies, highlighting the personal, societal and legal risks posed by this rapidly advancing technology. 

[1]Today, Deep Fakes have become buzzword for face-swapping technology. It refers to the manipulation or fabrication of audio, video, images and other signals through AI and machine learning techniques which makes use of deep neural networks (DNNs), generative adversarial networks (GANs), Variational Auto Encoders and sophisticated algorithms to create hyper-realistic synthetic media that can impersonate anyone with alarming accuracy. Synthetic Media refers to the media partially or wholly generated by the computers which includes deepfakes. The prevalence of this technology has progressed enough to raise concerns. For instance, deepfake pornography has violated personal boundaries of many women, celebrities like Scarlett Johansson and Gal Gadot being among the very first victims. Deepfakes of political figures like that of Barack Obama’s which was posted by Buzzfeed in 2018 showed the frightening realism with which deepfakes can mislead the public by altering speeches and showcasing fake events. Financially, these are used in scams, as fraudsters impersonate CEOs or family members to gain financial benefits. The recent movie “CTRL”, starring Ananya Pandey, perfectly depicts the potential of this technology to scatter the life of the victim and how the world would be left exposed if no legal and ethical actions are taken at the right time.

As AI is taking over the world, it becomes imperative to critically assess the legal provisions at international and national levels whether the steps taken are enough to protect the individual rights or not. The blog however specifically focuses on legal framework concerning deepfakes in India and debates around moral and legal dilemmas in India. It concludes with the understanding that the present legal provision is not sufficient and would require major joint efforts of both technologists and decision makers to uphold the rights of the citizens.

The Technology: Can you and me detect it?

The most basic technique behind deepfake generation is GANs i.e. generative adversarial network which consist of two competing neural networks. The “generator" network builds pictures or videos within a GAN framework by learning patterns within an extensively large dataset, while the "discriminator" network critiques these results in a bid to differentiate between the generated material and actual dataset. It keeps improving and enhancing the propensity for producing outputs until you reach the stage that the discriminator is finally deceived by the produced media. This process is very similar to an artist creating a portrait where he enhances the portrait through his skills(generator) until he is able to deceive the judge’s eyes(discriminator), thus successfully leaving the audience in awe (viewer).

A recent study published in Proceedings of the National Academy of Sciences USA, says “synthetically generated faces are not only photo-realistic, they are almost indistinguishable from the real thing and are considered more reliable”. But identification is still not completely out of question. Paying attention to details like number of image flashes, incongruities between proportion of body and face, mismatched lip movements, blinking of eyes, details inside the mouth etc. help to detect deepfakes.

Is it so simple? Not really. You might need to consider developing tools for its detection. The Deepfake Detection Challenge (DFDC) hosted by Kaggle is a significant effort towards creation of detection tools awarding a prize of $1,000,00 to winners supported by tech giants like AWS, Facebook, Microsoft, and the Partnership on AI.Companies like YouTube recently acknowledged the requirement for the creation of powerful tools for content creators to ensure online protection like 'synthetic-singing-identification technology,' which will allow creators to identify imitated voices.

Talking about any recent steps taken by India to combat this problem recently would include the efforts taken as a part of India AI Mission by MeitY where it funds to address problems like deepfakes. According to its proposal, the government is looking out for tools for real time deepfake detection, marking AI generated content which is traceable and developing ethical frameworks to promote fairness. It is further allocating ₹20.46 crore under the safe and Trusted AI pillar.

Are the legal provisions in India enough to tackle these problems?

Imagine a political leader giving inflammatory speeches. Wouldn’t that spread like wildfire in a country like India? For this Rule 3(1)(b) of the IT Rules mandates due diligence, however enforcement remains poor thus making India vulnerable to misuse of deepfake content. While other rules such as Section 66D (identity theft) and Section 66E(privacy) of IT Act provide indirect protection against defamation, identity theft and privacy violation and copyright infringement, deepfake regulation remains largely unsatisfactory.

Some of the recent cases like Anil Kapoor vs. Simply Life India & Ors. and Jaikishan Kakubhai Saraf Alias Jackie Shroff vs. The Peppy Store &Ors as 'Bhidu represent the growing awareness of the issue where they sought to get relief against the misuse of their likeness. Further, IT Rules 2021 instructs platforms to identify and delete such content within 36 hours to retain their safe harbour protections.

None of the laws and provisions, however are particular or impactful enough to tackle the challenges presented by Deepfakes. India thus needs to take inspiration from the international AI laws to bring some impactful changes in the rapidly developing AI world.

What are the international provisions?

In the United States, laws require Deepfake content to have labels or watermarks so that people know what is real and what's fake. Some states, such as California, even make it illegal to use Deepfakes to influence elections. Imagine if fake videos of leaders giving inflammatory speeches surfaced during an Indian election—clear labelling could prevent chaos! China has gone one step ahead by demanding that all Deepfake content must be labelled and the onus on platforms is to prevent such fake videos from causing damage. Their rigid approach reflects how labelling and proactive regulation can control this technology. The European Union has rules for platforms like Facebook and YouTube forcing them to clarify how they actually handle Deepfakes. Non-responsible moves by them cost them heavy penalties. India needs to follow something similar to its own advantage regarding the removal of harmful content soon.

Learning these lessons, it can create strong legal frameworks that defend people against fraudulent videos while, at the same time, creating trust in this digital world.

References

1-Vaibhav Agrawal, 'Ethical Implications of Deepfake Technology in User Interfaces' (2024) https://www.researchgate.net/publication/377413712_Ethical_Implications_of_Deepfake_Technology_in_User_Interfaces accessed 14 November 2024.

2-Sarah Jamison, 'Ethical Implications of Deepfake Technology in the Context of Pornography Through a Kantian Perspective' (2024) https://www.researchgate.net/publication/383040216_ETHICAL_IMPLICATIONS_OF_DEEPFAKE_TECHNOLOGY_IN_THE_CONTEXT_OF_PORNOGRAPHY_THROUGH_A_KANTIAN_PERSPECTIVE accessed 14 November 2024.

3-Synthesia, 'The Future of Synthetic Media' (Synthesia Blog, 2024) https://www.synthesia.io/post/the-future-of-synthetic-media accessed 14 November 2024.

4-TechUK, 'Synthetic Media: What Are They and How Are TechUK Members Taking Steps to Tackle Misinformation and Fraud?' (TechUK, 2024) https://www.techuk.org/resource/synthetic-media-what-are-they-and-how-are-techuk-members-taking-steps-to-tackle-misinformation-and-fraud.html accessed 14 November 2024.

5- Rishita Yadav, 'Title of Article' (2024) https://jlrjs.com/wp-content/uploads/2024/02/27.-Rishita-Yadav.pdf accessed 9 January 2025.

6- Rouse, 'AI-Generated Deepfakes: What Does the Law Say?' (Rouse, 10 January 2025) https://rouse.com/insights/news/2024/ai-generated-deepfakes-what-does-the-law-say accessed 9 January 2025.