Artificial Intelligence or AI has always been a topic of discussion due to its impact on human life. Initially, majority of the reasons why AI used to be in news were positive reasons, that is, those reasons which stressed on how AI has been successful in making human life more “plain-sailing” ; and no doubt, these positive reasons still remain a topic of discussion but, as of recent, majority of the discussions surrounding AI have made a swift shift towards the concerns that are slowly yet definitely emerging as a result of invention of certain types of new AI techniques which are slowly yet surely encroaching the moral, social and even legal sphere of the society.
This research paper stresses on one such new technique of AI that has been successful in raising concerns across the globe, namely, deepfake technique. It discusses both the problems and solutions related to deepfake technique, and also the historical development of deepfakes by taking into cognizance various notable incidents which succeeded in garnering mass attention to the implications of AI generated techniques. This research paper focuses on the legal perspective of deepfakes and AI by scrutinizing various laws that are related to deepfakes.
KEYWORDS
Artificial Intelligence, Deepfakes, Logic Theorist, Deep learning technology, Generative Adversarial Networks (GANs), Information Technology Act,2000.
INTRODUCTION
Artificial Intelligence or AI has always been a part of human life back from 1956 followed by the invention of Logic Theorist. Logic Theorist was a program primarily based upon the problem-solving skills of humans and is, therefore, often regarded by many to be the first artificial intelligence program; It was presented for the first time at Dartmouth Summer Project on Research Artificial Intelligence (DSRPAI), 1956.
Artificial Intelligence refers to the capacity of a machine to successfully execute those functions that are typically considered to be related with the functions of the human mind. Such as, providing solutions to a problem, carrying out artistic work, composition of music etc. After the invention of Logic Theorists and its subsequent presentation at Dartmouth, Artificial Intelligence emerged as a subject that was intensively researched upon by many innovative minds. As a result, AI flourished by leaps and bounds in the coming years; the result of which is quite apparent in the present world.
In the present world, when we look at our surroundings, we can discern how evident AI has become in our day to day lives. We now possess the ability of collecting huge sums of information that would have been otherwise too cumbrous for a human with average aptitude to process, hence, this age is also referred to as “Big Data” Age. In this age, we can see AI being applied across various sectors such as marketing, banking, entertainment and medical. From having Siri in mobile phones in 2011 to having ChatGPT by OpenAI in 2022, we have come a long way. Every now and then, when we open newspapers or browse on e-devices, we can always find one or more articles related to advanced technologies emerging as a result of AI.
As of recent, one such new technology related to AI which is going to be the major focus of this research paper, is making rounds on various discussion platforms across the globe and has become quite infamous in majority of the Indian households because of the very much talked about Rashmika Mandanna deepfake video which featured a famous Indian actress entering a lift. An even more recent instance of deepfake occurred on Jan. 15, 2024, where Sachin Tendulkar, renowned cricketer was seen advertising a gaming app. The similarity in both of these videos is that the events depicted in these videos only happened in the target video and not in the physical world. All this was made possible due to a rather new technology of AI known as “deepfake”.
This research paper aims to introduce the readers with deepfakes and AI, how deepfake is generated, notable incidents of deepfake and types of deepfakes. This research paper also takes into cognizance the legal repercussions of deepfake, various laws related to deepfake, notable case laws, thereby giving suggestions to control the illegal use of deepfakes.
RESEARCH METHODOLOGY
This paper aims to vividly look into the concept of deepfake by comprehensively analysing and studying various pre-existing information available across the internet in relation to this topic. Hence, making its methodology purely doctrinal in nature. This research paper is made after compiling and synthesizing already existing articles, blogs and dissertations in relation to the concept of deepfakes and its legal implications and therefore, also involves comparative analysis of pertinent legal framework of different countries in relation to artificial intelligence.
REVIEW OF LITERATURE
Literature reviewed for this research paper is as follows:
Rockwell Anyoha in her Harvard special edition article titled ‘The History of Artificial Intelligence’ writes about the historical development including both the success and setbacks in the formation of AI, the present scenario and the future of AI.
Jia Wen Seow, Mei Kuan Lim, Raphael CW Phan & Joseph K. Liu in their article titled “A comprehensive overview of Deepfake: Generation, detection, datasets, and opportunities” comprehensively writes about deepfake generation, opportunities and research trends in both deepfake generation and detection.
Arash Heidar from Department of Software engineering, Halic University, Istanbul, Turkey in his advanced review titled “Deepfake Detection Using Deep Learning Methods: A Systematic and Comprehensive Review” for WIREs data mining and knowledge discovery discusses about the basic concepts and terminologies related to deepfake, Deepfake detection, fake image detection, fake video detection, attribute manipulation, expression and identity swap.
Also, a major portion of the legal perspective is based on the comprehensive understanding of various Sections present in the “Information Technology Act, 2000” , as released by the Ministry of Electronics and Information Technology.
WHAT IS DEEPFAKE?
The term “Deepfake” is made by the combinations of two words, namely, deep + fake. When used in the context of the term deepfake, the term deep symbolises AI deep learning technology, that is, the type of machine learning technology used in the creation of deepfakes. On the other hand, the term “fake”, as is evident, symbolises that the content generated by the use of deepfake technology is not real. According to Britannica, the term Deepfake was coined in 2017 and came to use when a moderator on Reddit created a subreddit titled “Deepfakes” where he posted pornographic content with the faces of celebrities inserted into the existing pornographic videos.
HOW DEEPFAKES ARE MADE AND THEIR DIFFERENT TYPES.
As previously discussed, deepfakes are made with the help of AI deep learning Technology but it is not the only type of technology used. The other type of technology that is used for the creation of deepfakes is Generative Adversarial Network technology.
Deep learning technology is that technology in which algorithms that are made with an intent of mimicking the structure and function of the brain are used, known as Artificial Neural Networks. They are used to process and analyse large amount of data available. Deep learning technique finds its uses in computer vision, network language processing, speech recognition and robotics.
Whereas, on the other hand, Generative Adversarial Networks (GANs) technique is that type of deepfake technique which uses two neural networks, namely, generator and discriminator. The generator, as the name suggests, has the role of generating various deepfakes. On the other hand, the discriminator, as the name suggests, has the duty of discerning whether the deepfakes so generated are actually real or not. The generator continues to create different deepfakes to be viewed by the discriminator and continues to do so until and unless the discriminator approves those sample as being real. As a result, in the end, the generator produces highly realistic synthetic data which can easily fool a human who may get an access to the said data.
Deepfake technology’s main aim is to create such types of synthetic media which appears to be highly realistic and completely mimics the target (usually, a real person), with the aspects of the content being manipulated in such a manner that the result completely persuades a person with average acumen into believe in the reality of the content so produced.
As a result of the above-mentioned techniques, different types of deepfakes may be successfully created. Due to the increase in the number of different types of deepfakes that are being produced nowadays, certain classifications have started emerging among deepfakes on the basis of what has been changed or created by the use of deepfake. Some of the types of deepfakes so classified are as follows:
- Face Swap: In this type of deepfake, the face of one person is swapped for the face of another person with the help of techniques used for generating deepfakes.
- Attribute Editing: In this type of deepfake, the attributes of a person are changed and therefore, often results in him looking quite different virtually, from how he looks in the real world. Examples of attribute editing: change in height of the target, change in hair colour, eye colour etc.
- Face Re-enactment: In face re-enactment, unlike face swapping, in which the face of the person in the video is completely changed with the face of the target person, it is only the expressions of one person that are swapped with the other person in the video.
- Fully Synthetic Material: Fully Synthetic Material is created by the type of the technique that is used by the very famous ChatGPT created by OpenAI. In this type of technique, there is no tempering with the original video or photo, but instead it deals with the creation of such synthetic material, which does not exist in the real world but only in the virtual world. Therefore, a fully synthetic material is created when the material is created solely by the use of AI and there is absence of any previous video or photo on which the AI was commanded to work upon.
- Audio Swap: In audio swap type of deepfake, the audio of one person is swapped with the audio of the target in the target video.
NOTABLE INSTANCES OF DEEPFAKE
The first deepfake traces its roots back to the late 2017s, when the face of Gal Gadot, a famous Israeli actress and model had been placed on an existing pornographic video, which made the viewers think that it was actually the actress who was engaged in the acts so depicted in the video.
After this first incident of deepfake, the deepfake technology ended up developing by leaps and bounds, making some well-known deepfakes which left their imprint in the virtual world. Some of the notable examples of deepfakes following the reddit incident of deepfakes are as follows:
- Nov. 17, 2018, Jim Acosta doctored Video: This famous video that was distributed by the Trump administration in 2018 was not a real video but was actually a fully synthetic material deepfake. It is alleged that this video was distributed by the Trump administration to justify the banning of Mr. Jim Acosta, a CNN Reporter: and therefore, as a result, in this video Mr, Acosta was made to look rude and aggressive.
- April 9, 2019, Anti-Malaria Deepfake: David Bekham, an English Footballer featured in a video which was distributed on April 9, across various social media pages, in which he was seen giving a message to end malaria in nine different languages. It is alleged that the video was created using puppet deepfake technique. Due to the distribution of this video, a new fear emerged in the minds of the masses about the potential use of deepfake technology by the politicians or newsreaders for the fraudulent purposes.
- May 22, 2019, Pelosi Doctored Video: Nancy Pelosi, former speaker of the United States House of Representative was made subject to the technique of deepfake in 2019 by the distribution of a synthetic media which gave the viewers the impression of Pelosi being intoxicated or otherwise impaired.
- June 7, 2019, Mark Zuckerburg Deepfake: The CEO of Facebook was also not able to escape the hideous world of deepfake as a quite controversial deepfake of Mark Zuckerburg was also made and subsequently distributed across the web in 2019. This deepfake of Mark Zuckerburg was quite different from other humorous deepfake videos available across the internet, as in this video, Mr. Zuckerburg with the help of mouth swap deepfake technique was seen approving all the allegations of data breach on Facebook.
- February 18, 2022, Zelensky Deepfake: The deepfake featuring Ukrainian President Volodymyr Zelensky was telecasted on a hacked Ukrainian Channel. In the video the president of the war-torn Ukraine was heard informing the viewers that the Ukrainian military has surrendered and as a result, the war is over; raising new concerns as to the use of deepfakes in warfare practices of propaganda and rumour.
Apart from the above mentioned noteworthy instances of the use of AI there are other famous instances too in which the deepfake technology was used, which includes, the Pope Francis puffer jacket deepfake, a deepfake video of the arrest of Former US President Donald Trump and previously discussed Rashmika Mandanna video.
LEGAL PERSPECTIVE OF DEEPFAKE TECHNOLOGY
From the various examples cited in the preceding paragraph, we can easily perceive how deepfake technology can be used illegally and has the potential of threatening the legal rights of people who are being involved and subjected to deepfake. We need to specifically understand the legal repercussions of deepfakes, current laws that can be used to curb illegal practices by the use of deepfakes and some notable case laws; which will eventually lead us to understand the urgent need of enacting laws related to deepfakes in India.
- Legal Repercussions of Deepfake Technology
A nimiety of possible crimes exist that may be caused by the use of deepfakes, more easily. Without prejudice to the vast possibility of crimes that may occur due to the use of the deepfakes technique, some of the implications are as follows:
- Identity Theft: Identity theft refers to that fraudulent act in which a person intentionally ends up obtaining the identity of another person by obtaining the personal identifying information of the latter and then uses the same to gain unjust benefit from the same. The use of deepfakes to commit Identity thefts can lead to false representations of individuals and therefore, can also be used to manipulate public opinion, thereby, causing harm to the reputation and credibility of the individual whose identity was so stolen.
- Practices affecting elections: As has been previously mentioned, David Beckham, a well-known footballer, had used deepfake in his video, so that it appears to the viewers that he spoke nine different languages while filming the video. This incident successfully made the people consider the possibility of the use of Deepfake by politicians to garner public favour by spreading false or misleading information about the qualifications of political candidates and therefore, influencing the outcome of elections. In a country like India, such use of deepfakes will ultimately end up undermining the integrity of India’s democratic process.
- Data Privacy & Obscenity: As was seen in the first instance of deepfake that took place in late 2017s on Reddit, we can see that due to the use of deepfakes and AI, a person’s attributes can be given to another person and therefore, depicting the target person to be doing certain acts which the person in the original video might be doing. This raises a serious concern as to the violation of privacy of an individual and diminishing the image of the target in the society.
Aside from the Legal implications mentioned above, certain other legal implications can be inferred from the various notable incidents of deepfakes that have been previously mentioned.
- Notable Lawsuits Against Deepfake in India
Although there are various cases in India that are related to deepfakes, there are few lawsuits that specifically paved the way against deepfakes in India by making the masses more aware about deepfake technology.
- Anil Kapoor v. Simply Life India & ors.: In the following case, the famous Bollywood actor Anil Kapoor had filed a case against the defendants after discovering AI generated deepfake content which used plaintiff’s likeness and voice.
- Amitabh Bachchan v. Rajat Negi & ors.: In the following case, the legendary film actor Mr. Amitabh Bachchan was granted injunction against the unauthorised use of his attributes and personality rights.
- Current Laws in India
As of now, India doesn’t have a legislation that specifically deals with deepfake technology; hence, keeping a mist around the remedies which are available to a person whose legal right has been infringed by the use of deepfakes.
However, just because India doesn’t have a specific law for deepfakes does not mean that any person who finds his legal rights to be infringed due to the use of AI technologies cannot go to claim remedies at courts. Section 66E of the IT Act of 2000 is applied in cases related to deepfake offences which involve the capture, publication, or transmission of a person’s photographs in mass media, leading to the infringement of his privacy. If a person is booked under Section 66- E and it is proved in the court of law that he indeed committed such an act which is prohibited under section 66-E, then he shall be imprisoned for a term which may extend to a time period of three years or with a fine of not more than Rs. 2,00,000 or in certain cases, both.
Aside from this, Section 66-C of IT Act, 2000 applies in cases where there is identity theft by the use of deepfakes. If a person is booked under Section 66-C and it is proved in the court of law that he indeed committed such an act which is prohibited under section 66-C, then he shall be imprisoned for a term which may extend to a period of three years and with a fine not exceeding Rs. 1,00,000.
Section 66-D of IT Act, 2000 applies in case of cheating being committed by the use of computer resources. If a person is booked under Section 66-D and it is proved in the court of law that he indeed committed such an act which is prohibited under section 66-D, then he shall be punished with imprisonment, the term for which may extend to three years and fine not exceeding Rs. 1,00,000.
Section 67 of IT Act, 2000 applies in cases of publishing and transmitting deepfakes that are of an obscene nature in electronic form. It has provision for imprisonment which may extend to a period of three year and with a fine not exceeding Rs. 5,00,000; if such a wrong is committed. However, If the wrong has been committed for a second time, the defendant shall be sentenced to imprisonment which may extend to a period of five years and with a fine not exceeding Rs. 10,00,000.
Also aside from the above-mentioned statues, the Ministry of Electronics and Information Technology issued an Advisory on November 7, 2023, related to deepfakes, with certain directions for social media intermediaries.
SUGGESTIONS AND CONCLUSIONS
As discussed above, there are various Sections in IT Act, 2000, which can deal with deepfakes. However, the mere existence of these sections does not fill the void that is there in our legal system with regards to deepfakes; and this void is felt time and again by the masses whenever a new notable incident of deepfake takes place.
Hence, it is of immense importance that specific laws should be enforced in India to deal with the rising cases of deepfakes or necessary revisions should take place in the IT Act to include deepfakes as a separate type of cybercrime. We should also remember that cases of deepfakes are not something limited to India but they can be found throughout the globe. Therefore, in this age of globalisation, joint cooperation between different countries of the globe, to tackle the repercussions of deepfake will make the implementation of deepfake laws more effective.
We should analyse various steps taken by different countries to tackle deepfake and try to embody them in our legal system. For example,
- The Indian government should pass a law which penalises the distribution of illegal deepfakes which have a potential of causing harm to public interest; like its South Korean counterpart.
- The Indian Government should try to clearly label deepfakes on social media to prevent public confusion; like its Chinese counterpart has stressed.
- The Indian Government should try to establish a task force especially to address deepfakes; like its US counterpart has advocated the Department of Homeland Security (DHS), to establish a task force to address deepfakes.
Hence, we can see that the whole world is currently laying stress on limiting the repercussions of AI & deepfake technology by taking the help of law. Therefore, it is high time for Indian Authorities to start taking definitive steps related to deepfakes to curb the illegal use of these modern technologies and to provide a safe virtual world to the masses.
KAASHVI MALIK
DR. B.R. AMBEDKAR NATIONAL LAW UNIVERSITY, SONIPAT, HARYANA