Name of author: Shruti Shrivastav (Final year student, Government Law College, Mumbai

INTRODUCTION

Deepfake and its tendency to trick viewers into believing in its originality has become an alarming concern in recent years. In 2023, the surfacing of a promiscuous video of actor Rashmika Mandanna sparked discussions on the potential misuse of deepfakes prevailing on the internet. Such misuse is antagonistic to individuals’ right to privacy, as it tends to manipulate their identity, face, and features. This makes it a grave concern, given that the right to privacy is a recognised human right under Article 12 of the Universal Declaration of Human Rights (UDHR). In India, it was construed as a fundamental right under Article 21 of the Constitution by the Supreme Court in the landmark case of Justice KS Puttaswamy vs. Union of India case. However, the violation of the right to privacy is just the tip of the iceberg i.e. the most apparent violation caused by deepfakes. Another inherent menace embedded in the usage of deepfakes is the infringement of copyright. Deepfakes employ audiovisual content as input, modify it as part of its processing, and generate an output that significantly differs from the original content. This means it manipulates the very content that copyright laws tend to protect. Copyright protection laws worldwide have not been able to keep pace with the boom in AI and the resultant use of deepfake technology. This demands an amendment in the existing laws or enactment of new laws to ensure that copyright protection does not fall prey to the usage of deepfakes in the long run.

DEEPFAKES AND AI

Artificial Intelligence, as the name suggests, is intelligence exhibited by machines by which they tend to showcase certain human qualities and perform tasks as directed. The term includes various applications like machine learning, deep learning, robotics etc. So vast is the ambit of AI that there is no strict definition attributed to it. It is understood as an umbrella term that encompasses any human behaviour or intelligence demonstrated by machines. One such subset of AI is deepfake. It is a type of AI that is used to create convincing fake images, videos, and audio recordings. These videos are made through a pair of algorithms that generate counterfeit media by swapping the original one with a fake one and then differentiating the two (original and fake) until no difference is found. The algorithm is immaculate enough in its task to make the human eye believe in its work. In conclusion, to say that if AI is a parent, then deepfake is its prodigy would not be an exaggeration.

DEEPFAKES AND COPYRIGHT INFRINGEMENT: USA vs INDIA

• USA

The menace of copyright infringement via deepfakes is relatively unattended and less spoken of worldwide. The same is reflected in the legal framework of the United States of America, where the current laws and judicial precedents have not only been insufficient in tackling the consequences of the same but have also given rise to legal quagmires in the name of ‘fair use’ doctrine. This doctrine finds place in Section 107 of the Digital Millennium Copyright Act (the “Act”), 1988 of the USA. It acts as a defence for using copyrighted work for commenting, criticizing or parody, for instance, quoting paragraphs from a news article for educational purpose. Under this section, a four-factor test is laid down to determine whether any usage of a copyrighted work would qualify as ‘fair use’. These factors are:

• purpose and character of the use
• nature of copyrighted work
• amount and substantiality of the portion of work used
• effect of the use on the potential market    

However, the Act does not provide any criteria for the application of these factors to determine fair use. Thus, fair use is determined on a case-to-case basis. In the famous case of Campbell vs. Acuff Rose, a subset of this doctrine was laid down, called ‘transformative use’. Nestling it within the first factor, i.e., purpose and character of the use, the US Supreme Court observed, “the more transformative the new work, the less will be the significance of other factors, like commercialism, that may weigh against a finding of fair use.”. A major repercussion of this doctrine has been the parallelisation of ‘transformative use’ and ‘derivative use’. This means that any work that uses a pre-existing work as its foundation and adds any supplement or inserts any change in it, is interpreted as fair use, no matter the distortion it brings about of the original work and the harm it causes to the subjects of that work. This was seen in the case of Patrick Cariou vs. Richard Prince, where Cariou’s images of people belonging to the Rastafarian community published after their consent in the Yes Rasta, were interspersed with colors and images of white women by Prince. Not only was the tediously crafted work of Cariou stripped of its copyright protection, but so were the people of the Rastafarian community, of their consent to publication, in the name of fair use or so-called transformative use. Deepfakes are analogous to such derivative use, wherein the AI morphs the original content, which is detrimental to the person so morphed and also to the owner of the content. This can lead to rampant misuse of the technology in the name of creating ‘transformative’ videos with a ‘different purpose and character’. Additionally, the fair use doctrine is open-ended and is not accompanied by fixed rules. This makes the US jurisprudence on what qualifies as fair use of copyrighted material vague and capable of multiple interpretations.

However, the Senate is no longer sweeping the alarming issue of deepfake under the rug with a new bill called the Content Origin Protection and Integrity from Deepfake Media Act (COPIED Act) being introduced that contains guidelines for detecting synthetic content and incorporating content provenance information. This would help prevent the usage of protected information in training AI models to generate content, thus providing greater control to the original owners over their content and promoting transparency. The bill, which also defines the term deepfake under Section 3(7) as ‘synthetic content that appears authentic and creates false impression’, is currently pending in the United States Congress.

• INDIA

India does not have any explicit law to deal with copyright infringement caused by deepfakes. The Copyright Act of 1957 contains provisions for copyright protection and thus, is currently the go-to legislation for all such matters. Section 52 of the Act enumerates the list of acts which are not considered copyright infringement. Section 52(1)(a) talks about ‘fair dealing’ of any work, which is one such act. Since this section provides an exhaustive list of acts that are not considered copyright infringement, it implies that any act not falling under the ambit of the said list would be considered copyright infringement. Deepfake does not form a part of this list, thus, the immediate conclusion is that deepfake is indeed copyright infringement. However, the absence of an explicit mention of deepfakes, (due to the technology being a recent trend, against the act that was passed in the late fifties) can give a pass to deepfake usage causing copyright infringement in the name of fair dealing. This predicament is similar to the one posed by the fair use doctrine in the US.

On the other hand, the complete dismissal of deepfake as copyright infringement would also be detrimental since it would obstruct the technology’s usage for productive purposes. Section 52(1)(a)(ii) talks about ‘criticism or review’ as part of fair dealing. Productive usage of deepfake would ideally find a place under this provision of criticism or review, however, the inclusion of deepfake within its ambit is not backed by any judicial precedent. Judicial remarks on deepfake have generally been in the context of its misuse causing privacy breaches, the aspect of copyright infringement by deepfake or its authentic usage under fair dealing is still to be touched upon.

CONCLUSION

It can be seen that the legal framework of both countries is not adequately equipped to deal with the rising trend of deepfakes, however, a striking difference that stands out is that the US regime is being revamped to tackle the same. This is reflected by the introduction of the COPIED Act in the Senate, which has been specially drafted to dissociate deepfakes from original content. No such law has currently been introduced in India.

Referring to the fair use doctrine of the USA, and the fair dealing doctrine of India, both are devoid of appropriate elaborate provisions and rules to keep a check on deepfakes. No mention of a trend as recent as deepfakes in the acts that were passed in 1988 and 1972 is understandable. However, it is time that provisions for deepfakes are incorporated into these acts to prevent any further misuse of the technology. To determine what could be termed fair dealing or fair use of copyrighted work, the legislature of both jurisdictions has the responsibility of extensive deliberation to expound upon the idea and come up with guidelines for the same. This should not only be limited to what should and shouldn’t be termed as copyright infringement but should also explain the ownership of copyright in the case of AI-generated content created using existing content. Whether such content would belong to the original creator, the machine (i.e. AI) or the person who directed AI to create the content using original content is a question that needs to be answered.