Deepfake Pornography: A Legal and Ethical Menace

Introduction

Ever came across the videos of political leaders delivering speech completely contrary to their ideology or a well-known actor doing bizarre things in public? If the answer is in the affirmative, then, there is a good chance that you have encountered a deepfake. While the term “deepfake” doesn’t have a globally agreed definition, it can be termed as an image, audio or video hoax created through Artificial Intelligence (AI) and Machine Learning (ML).

With the rapid growth and advancement of ML, deepfake has the ability to synthesize very realistic audio, video and image. Its ability to produce such convincing images/videos although amusing, also raises concerns about the misuses it can lead to. To name a few: misrepresentation and identity theft; misinformation against government; fabrication of evidence; obscenity/ breach of privacy amongst others.

In this piece, I shall be dealing with one of the most disturbing strains that we owe to the AI technology i.e., Deepfake Pornography (DP).

What is Deepfake?

DP or non-consensual pornography is an AI visual generating application where someone’s face is superimposed onto some other person’s body or placed in a situation where they were actually not. However, this technology is not new. In 2017, certain Reddit users produced synthesized porn videos of the celebrities with the help of AI.

But, with the new dynamics of communication network, the technology has become more abundant. It now uses a more competent deep learning architecture called generative adversarial network (GAN). GAN comprises of two neural networks- a generator and a discriminator. While a generator creates fake samples that resemble real data of the training set, the discriminator evaluates the synthetic images produced and provides the necessary feedback to improve the likeness with the real image. This process lasts until the generator brings about an indistinguishable image.

Deepfake: A Multilayered Issue

More often than not, DP is considered to be just a form of image-based abuse or digitized sexual assault where the perpetrator uses the image without the person’s consent to create a sexually explicit content. But the reality is, it has many latent layers attached to it which can have grave ramifications on physical, mental and social levels. Thus, it not only emanates fundamental right issues but also ethical and gender related issues.

1. Fundamental rights issues: DP stands in total contravention with the fundamental rights as the victim suffers from lack of control and agency over his/her own body. It intervenes with the right to privacy and the right to be forgotten, which is an integral part of right to life and liberty enshrined under Article 21 of the Constitution of India, 1950. This was highlighted by the honourable Supreme Court in the case of K.S. Puttaswamy. Here, the court postulated that the right to privacy also covers the right to be let alone. The court emphasized that ‘Thoughts and behavioural patterns which are intimate to an individual are entitled to a zone of privacy where one is free of social expectations.’

In the case of Subhranshu Rout v. State of Odisha, the Orissa High Court stressed on the importance of ‘right to be forgotten’ in revenge porn cases. The court said that the non-recognition of the same could be detrimental for the safety and protection of women against the blackmailing on the pretext of the non-consensual pornographic images.

2. Ethical issues: The application digitizes the real images of the people to create hyper-real avatars. The technology empowers the users to customize the avatars by choosing different features like skin colour, body shape etc. Thus, it is an AI tool which objectifies the body of women. It seems that this technology is built to reinforce the patriarchal notions that men own the women’s bodies. The rise of online misogyny can also be witnessed from the revenge porn where either of the partners share the intimate sojourn online/offline when the relationship turns sour. This sub-category is particularly more morally problematic as the offender puts the private pictures of someone, they have admired just to satisfy his vengeance.

3. Gendered issues: According to the research agency Sensity AI, at least 90% of the deepfake videos online are porn videos and 90% of those deepfake porn videos are of women. The sources suggest that these algorithm models are trained on women’s bodies. So even if one feed a male image into the face-swap application, it’s not going to provide such realistic results. This is how AI biasness comes into the picture. The algorithms also get prejudiced in weighing people and situations. This biasness is generally the result of the training data fed by humans, which is subject to the biased human thoughts and the societal norms.

Laws Around the Globe

Deepfakes have garnered widespread attention due to its severe implications. Consequently, the countries have come up with necessary legislations. However, while legislating, the lawmakers have faced three-fold challenges- Firstly, the identification of deepfakes is quite hard as the advances in AI and ML have made it easier to produce naturalistic content. Secondly, there is a problem of the victim recognition. Unlike, other crimes, here two victims are involved. One whose face is superimposed, and the other, whose body is used to create deepfakes. Thirdly, there is an enforcement problem as well. Given the far reach of internet, it is possible that the perpetrator is located outside the country/jurisdiction of the enforcement agencies. Thus, it becomes very difficult for the law enforcement agencies to trace the perpetrator. Additionally, it is quite unclear as to how domestic laws will apply on an international DP creator.

It is imperative to analyse as to how different jurisdictions are dealing with DPs through legislative/policy measures:

1. The United States of America (USA)- USA does not have a federal law prohibiting deepfakes. Though an attempt was made in the form of a bill called Defending Each and Every Person from False Appearances by Keeping Exploitation Subject to Accountability Act, it is yet to be put for voting. The proposed bill entails certain pre-requisites for the producers of the deepfakes like digital watermarks, disclosure requirements etc. It creates both civil and criminal liability for the creation of deepfakes.

At the same time, few states like Virginia and California have enacted legislations. But these legislations suffer from significant loopholes. The legislations only prohibit the sharing of DPs; while there is no prohibition on mere creation of the DPs. This is quite startling as even making non-consensual pornography amounts to grave violation of the rights of the women. Additionally, Virginian law even requires an additional proof of malice on the part of the creator.

2. The United Kingdoms (UK)- The Online Safety Bill is one of the only legislations around the world containing provisions penalizing DPs. The bill was first proposed in April 2019 and is expected to become a law this year. The legislation was made with the intent of making UK ‘the safest place to be online’. The Bill states that the perpetrators will face severe prosecution irrespective of the fact whether the creator intended to cause distress or humiliation to the victim. However, similar to Virginia’s and California’s law, it does not prohibit the mere creation of DPs.

3. China- In 2022, the Cyberspace Administration of China (CAC), the Ministry of Industry and Information Technology (MIIT) and the Ministry of Public Security (MPS) jointly issued the Provisions on the Administration of Deep Synthesis of Internet-based Information Services. The provisions provide a comprehensive definition of ‘deep synthesis technology’ and include techniques like face generation, face replacement, gesture manipulation. Under the regulations, deepfakes cannot be produced without the users’ consent and those which are created should be clearly labelled as ‘edited.’ The provisions are applicable on two entities- the data synthesis service provider i.e., the company offering deep synthesis services and the deep synthesis service users who create, duplicate or publish synthesized content.

4. South Korea- South Korean law was proposed in the light of the fact that Korean female K-pop stars are at very high risk of being weaponized through non-consensual deepfake pornography. In 2020, the South Korea revised the Act on Special Cases Concerning the Punishment, Etc. of Sex Crimes and added provisions which restrains both creation and distribution of visuals which show them in sexually explicit content. Moreover, unlike Virginian laws in the USA, the existence of malice is not a pre-requisite. The perpetrator cannot shun away from his liabilities even if he contends that he created/ shared the content out of admiration for the subject and not out of malice.

5. India- There is no distinctive law in India which deals with deepfake technology or DP specifically. However, there are certain provisions in the Indian Penal Code, 1860 (IPC) and the Information Technology Act, 2000 (IT Act) which can create liability against the perpetrators. Section 66-E (punishment for violation of privacy), Section 67-A (punishment for publishing or transmitting of material containing sexually explicit act, etc. in electronic form), Section 67-B (punishment for publishing or transmitting of material depicting children sexually explicit act/pornography in electronic form) of the IT Act are relevant. Sections 292 and 294 (punishment for sale etc. of obscene material) of the IPC may also be referred. Additionally, Section 4 (prohibition of publication containing indecent representation of women) of the Indecent Representation of Women (Prohibition) Act, 1986 is also pertinent. Sections 13, 14 and 15 of the Protection of Children from Sexual Offences Act, 2012 can also be invoked to protect the rights of women and children.

The Way Forward

Present is the age of internet. Deepfake technology has filled the internet with non-consensual pornographic material. The technology raises an underlying question of ‘consent’ as DP makers treat women merely as an object for sexual gratification. The situation raises an immediate cause for alarm. The Ministry of Electronics and Information Technology will soon come up with the Digital India Act, 2023 that will replace the Information Technology Act (IT Act), 2000. The new Act is being brought to address the ever-evolving digital space. The Act will cover various aspects of ‘Online Safety and Trust’ including deepfakes, data protection etc.

Drawing inspiration from the legislations around the world, the following guidelines/ provisions should be incorporated in our domestic laws in order to curb DPs. Firstly, one can regulate the training data sets that is fed into the AI system. One can remove extremely explicit content and desist generator’s ability to produce damaging content. Secondly, one can instil the principle of transparency. This can be done by establishing a review mechanism of the content created, implementing real-identity information authentication so that the users producing false content can be tackled. Also, there is a need of periodical evaluation of the algorithms so that that the technical security can be ensured. Thirdly, appropriate forensic techniques should be developed to detect deepfakes. Besides that, content labelling in the form of tags, watermark should be mandated. Fourthly, the pre-requisite of intent should be abrogated as regardless of that, the bodily autonomy of the victim is impeded while creating DPs. Fifthly, it must be ensured that not only the sharing but also the making of DP is prohibited.

A strict law/ regulation is the need of the hour so that one can unleash the useful side of AI generated deepfake technology.

(This article has been authored by Janvhi Rastogi, a 5th Year student at Faculty of Law, Jamia Millia Islamia)

CITE AS: Janvhi Rastogi, “Deepfake Pornography: A Legal and Ethical Menace” (The Contemporary Law Forum, 16 October 2023) <
tclf.in/2023/10/16/deepfake-pornography-a-legal-and-ethical-menace/> date of access.