Artificial Intelligence (AI) has been at the forefront of development in numerous industries recently, including the legal spectrum. Its impacts have been both positive and negative.. Among these, the proliferation of deepfakes stands out as a critical issue due to its potential to undermine human rights.
The term ‘deep fake’ has been defined under Article 3(44)(b)(l) of the European Union Artificial Intelligence Act (EU AI Act) as ‘AI-generated or manipulated image, audio or video content that resembles existing persons, objects, places or other entities or events and would falsely appear to a person to be authentic or truthful’. A cursory examination of the definition reveals the considerable influence wielded by deepfakes and the potential for this technology to have adverse impacts. Effectively addressing the ethical concerns surrounding AI-powered deepfakes requires a multi-faceted approach encompassing technological advancements, public awareness, the establishment of clear legal regulations, and international cooperation.
Freedom of Expression
Article 19 of the International Covenant on Civil and Political Rights (ICCPR) and of the Constitution of Pakistan enshrine the freedom of expression. The use of AI-generated content, including the use of deepfakes, falls under the freedom of expression. However, it is essential to acknowledge that the freedom of expression under international and domestic law is subject to certain conditions, such as considerations of public order, decency, morality, or prevention of incitement to an offence. Thus, any law that limits such expression must be justified accordingly.
The harmful use of deepfakes goes beyond creating fake content to fulfil fantasies: it also includes tactics like harassment, intimidation, and silencing people. This creates an environment where fear and instability become prevalent. Such actions can significantly impede freedom of expression and inhibit participation in public discourse, especially among vulnerable groups and individuals. This chilling effect underscores the importance of addressing the misuse of deepfakes to safeguard the rights of all individuals to express themselves freely and engage in public life without fear of malicious manipulation or reprisal.
One such tool is revenge porn, which is disseminated with the intent to exact revenge or retaliate against an individual. For example, Rana Ayyub, an Indian journalist known for her vocal criticism of the Modi administration and the Hindu nationalist movement, was subjected to harassment as adversaries fabricated pornographic material featuring her and distributed it widely online. Such technology is known to promote gender-based violence and possibly lead to an increase in honour killing in certain communities. Thus, the unchecked use of deepfakes may result in implications on the right to life and the security of a person which is guaranteed under Article 6 of the ICCPR and protected under Article 9 of the Constitution of Pakistan.
Right to Privacy and Dignity
Consequently, the uncontrolled use of deepfakes has implications on the right to privacy and dignity. Article 17 of the ICCPR protects the right to privacy and prohibits unlawful attacks on a person’s honour and reputation. Dignity and privacy have also been protected under Article 14 of the Constitution of Pakistan.
Deepfakes present a substantial threat to the right to privacy due to their potential use in creating images, videos and audio that depict individuals participating in certain activities that are considered harmful, inflammatory or undignifying. These can include non-consensual pornography, manipulation of news, fraud or even identity theft, resulting in severe reputational damage. This misuse not only violates the right to privacy and dignity but also has profound consequences for victims in their personal and professional lives.
It is not surprising that deepfake pornography videos amount to 98% of all videos created by deepfake. Deepfakes facilitate the creation of scenarios and fantasies that may not be feasible within conventional pornographic content. This assertion is substantiated by data indicating that 94% of the aforementioned videos prominently feature individuals in the entertainment industry. A recent example concerns the pop singer Taylor Swift, who found herself a victim of deepfake images in January 2024.
Similarly, politicians, such as Italian Prime Minister Giorgia Meloni, have also fallen victim to this technology. The Prime Minister is now seeking €100,000 in damages for deepfake porn videos portraying her were published on the internet. Noelle Martin, an Australian advocate and attorney who experienced the harm of such illicit videos herself, emphasised that deepfake pornography can have enduring repercussions on a person’s reputation, dignity, career prospects, and personal connections. Moreover, she attributed her inability to obtain a job to this particular incident. Such videos can tarnish a person’s reputation, leading to stigmatisation and negative perceptions by employers, colleagues, and the public.
Right to Fair Trial and Due Process
Deepfakes can be used to fabricate evidence and manipulate witness testimonies, thereby posing a serious threat to the principles of fair trial, a right granted under Article 14 of the ICCPR and Article 10A of the Constitution of Pakistan. States all over the world allow digital evidence, including videos and audio. However, as technology continues to advance, this form of evidence is now more frequently coming under heightened scrutiny and public focus.
Consequently, the heightened risk of wrongful convictions underscores the pressing need for robust safeguards within the legal system to mitigate this emerging threat to the justice system. Moreover, ensuring the authenticity of digital evidence introduced in a case necessitates seeking expert opinions, a step that may inadvertently prolong the proceedings, contrary to the current trend of expediting the judicial process. Marie-Helen Maras & Alex Alexandrou interestingly highlight in their paper that digital evidence was introduced to corroborate oral evidence; however, with deepfakes existence, oral evidence will be needed to corroborate digital evidence (p. 259). Thus, deepfakes pose a significant challenge to the admissibility and authenticity of evidence in legal proceedings.
Right to Equality And Non-Discrimination
Article 25 of the ICCPR and of the Constitution both ensure the right to equality and non-discrimination of all individuals. Deepfakes can be misused in a manner that results in targeted attacks on specific groups or individuals on the basis of race, gender or religion, as well as other identity markers. This form of discrimination intensifies existing social tensions and reinforces harmful stereotypes.
An example of such a situation highlighted by Bobby Chesney and Danielle Citron is a scenario consisting of tense racial relations between a police force and a community. When a provocative deepfake video depicting a police officer using racial slurs or violence is circulated, the same could spark civil unrest or riots. Similarly, a reverse deepfake portraying a community leader advocating violence against the police could have equally dangerous outcomes. In a nation like Pakistan, where accusations of blasphemy can lead to severe consequences for individuals and even entire communities, as tragically demonstrated in 2023, when a mob, incited by purported actions of two Christians, engaged in looting and burning of houses and churches, the dissemination of altered videos can spark widespread uncontrolled violence
A recent example from Pakistan relates to the renowned Aurat March, an annual event conducted on Women’s Day to express solidarity with women’s rights, which has consistently encountered criticism from certain segments of Pakistani society. In 2021, a video from the March underwent manipulation, depicting the marchers chanting blasphemous content. The term ‘mullah’ was edited to appear as ‘Allah,’ portraying the marchers as calling upon God instead of referring to the clergy. This manipulated video was initially shared by three senior journalists before they subsequently removed their tweets. These actions further sparked hatred towards the March as blasphemy is considered a capital offence and one that often results in mob violence. The increasing accessibility of content editing through deepfake technology raises significant concerns about the potential escalation of such issues. This highlights the critical need for content authenticity verification to mitigate the risks associated with deepfake content.
Existing Legislation Regulating Deepfakes
It is evident that effective strategies must be devised to mitigate the harmful impacts of deepfakes. Article 52 of the EU AI Act states that providers of AI technology creating or altering image, audio, or video content to produce deepfakes must openly acknowledge that the content has been synthetically generated or modified. In California, the distribution of deepfake content featuring political candidates within 60 days of an election was prohibited by AB 730 20010. (a), a law that expired on January 1, 2023. Concurrently, California enacted AB 602, which banned the creation and distribution of pornographic deepfakes without consent.
Victims can leverage existing copyright, privacy, and harassment laws in their respective countries for recourse. For instance, section 188 of the UK’s Online Safety Act, 2023 prohibits the unauthorised sharing or threat of sharing digital media. In Pakistan, section 18 of the Prevention of Electronic Crimes Act, 2016 (PECA) specifically addresses acts that harm a person’s reputation or privacy, with penalties of imprisonment for up to three years and/or a fine of up to one million rupees.
However, a significant hurdle in providing grievance redressal for the misuse of depfakes is the identification of the abusers. These individuals often operate under the cloak of anonymity and disseminate their creations across international online platforms without regard for borders. Only a handful of countries have enacted legislation regulating AI, and even that legislation may not be strong enough to deter the negative impact of AI and deepfakes. It may take a few years – at the very least – for these countries to introduce robust laws tackling the advent of this technology.
Fighting Against Deepfakes
While countries are in the process of introducing laws regulating the use and abuse of deepfakes, victims of this technology need remedies that are easily and readily accessible. The initial step involves gathering evidence, such as taking screenshots or downloading the relevant file(s) to be submitted to grievance redressal mechanisms.
Various organisations are offering services and resources to have deepfake content removed from the internet upon a request from an individual. Tech giants such as Google have also been removing such content upon users reporting the same. Victims of deepfake videos can utilise the US Digital Millennium Copyright Act, 1988 (DMCA) to request the removal of these videos by citing copyright infringement. Google has streamlined this process for victims of ‘involuntary fake pornography’ by providing a separate form dedicated to such removal requests. This simplifies the procedure, allowing victims to submit their requests efficiently and effectively for the removal of harmful deepfake content.
Meta Platforms has implemented an online tool called ‘Take it Down‘ to specifically target and remove media depicting minors on Facebook and Instagram. This tool works by digitally assigning a unique fingerprint to the media, which is then shared with partner organisations. These partners then take action to either remove the content entirely or limit its further distribution, thereby enhancing efforts to protect minors online and combat the spread of inappropriate content.
OpenAI, the developers of Chat-GPT, has implemented limitations on their image-generating AI, DALL·E 2, to prevent the creation of violent, hateful, or adult imagery by excluding explicit content from its training data. Furthermore, they have utilised sophisticated techniques to avoid the realistic rendering of faces, including those belonging to public figures. Similarly, TikTok has taken a strict stance by prohibiting synthetic media featuring the likeness of private individuals and extending this prohibition to public figures if the content is used for endorsements or breaches policies related to hate speech, sexual exploitation, or harassment.
The government should engage in close collaboration with technology companies to effectively address the challenges posed by deepfake content on online platforms. This collaboration is paramount in developing and implementing technical solutions capable of detecting and mitigating the harmful effects of deepfake media. Standardisation, certification, policy guidance, and fostering public-private partnerships further enhance these efforts, ensuring a coordinated and comprehensive approach to combating deepfake content.
Above all. initiatives aimed at promoting media literacy and critical thinking skills can help educate individuals about the influence of deepfake content. For example, Finland has been combating the spread of fake news by education and has subsequently topped the Media Literacy Index. By preparing people to effectively distinguish between authentic and fabricated media, States can mitigate the harmful effects of deepfakes at the grassroots level.
Conclusion
In conclusion, the rise of deepfake technology poses a serious threat to privacy, reputation, fair trials, freedom from intimidation, equality, and non-discrimination. Addressing these complex challenges requires a holistic approach that integrates technological advancements, legal frameworks, international partnerships, and widespread public education. By proactively combating the misuse of deepfake technology, societies can reaffirm their commitment to upholding human rights principles and preserving democratic values in an increasingly digital age.
Centre for Human Rights (CHR) blog
