[This piece has been authored by Tanvi Agarwal and Tejaswini Sahoo, students at the West Bengal National University of Juridical Sciences, Kolkata.]
Introduction
If you are wondering how excellent the camera on your new brand phone would be, pay attention to what the developer says about AI. As technology advances, innovative ways of producing images emerge. Some are automatic and operate when a human photographer is not present. The moot question presently arises in the context of expanding AI and granting copyright protection to the images generated by AI.
The functioning of the AI is such that a large number of images are fed into the software. Subsequently, textual prompts or algorithms provided by the user scan the snippets of images of existing original artists and generate an innovative image in a creative way. However, legal trouble arises when these images are uploaded for commercial purposes as the controversy hovers around the rightful claim of authorship over these AI-generated images. The existing debate has been ignited pursuant to the release of Adobe’s new policy, which is speculated to circumvent this debate through its temporary revenue-sharing model.
Authorship rights do not vest with AI
Once the AI generates the images, the question of authorship arises and whether the AI software can be granted the same. The existing jurisprudence around the world indicates that authorship rights can only be conferred on a human entity. In Australia, Telstra Corp Ltd. V. Phone Directories Co Pty Ltd confirmed the centrality of whether copyright subsists when human intervention is too remote or ministerial. It confirmed the long-standing assumption that the author needs to be a “human”. The Irish Act also recognises the author as the person who creates a work. In the instance of computer-generated work, authorship is granted to the person by whom the preparations essential for the development of the work are done. Similarly, in the UK, the author is the person by whom the arrangements necessary for creating the work are undertaken. In the Indian jurisdiction, even though defined, the word “author” is left relatively open-ended in cases involving autonomously created computer work.
An analysis of §2(vi) of the Indian Copyright Act 1957 indicates that copyright protection is only provided to those computer-generated works which have been operated by human agencies.
The Delhi High Court affirmed this interpretation when copyright was asserted over a list created by a machine, and the Court dismissed the claim, citing, among other things, a lack of human interaction.
In furtherance of the lack of human intellectual abilities, AI is not recognised as a legal entity and fails to achieve the criteria of being a legal personhood. Moreover, to grant authorship, the work created must be original; the author must be able to exercise free and creative choices in producing the work. Due to the functioning of the AI, it can be argued that the image generated is a mosaic of the inputs fed into the software, indicating a lack of originality. Therefore, the AI software is not provided with authorship rights due to lack of human intervention, legal personhood and originality.
Escapism or a temporary model
As the AI software is not recognised as the author of AI-generated imagery, it opens a pandora of unaddressed rights issues. The dilemma is around the question of with whom the authorship rights of such AI-generated imagery can vest. While probing into this question, copyright law will have to ensure that it adequately protects the rights of the original artists of the work, developers of such AI software and the users of the AI-generated imagery. While copyright law juggles to obtain such a finely tuned balance, certain software developing companies like Adobe have found solutions outside the boundaries of the copyright system.
For instance, Adobe has recently recognised the technological progress made in terms of the creation of AI software and aims to embrace the technology after delving into the question of ethics. It recognises the unclarity in law, which may not favour granting authorship to the AI itself. In such a scenario, it aims to secure its user’s interests by protecting them against legal claims filed for copyright infringement. Therefore, while the law is not settled, it created a temporary model which permits the use of AI-generated images on their platform only after the developer of the AI software has obtained all the requisite licenses for usage of the work from the original artist of that work. This ensures the protection of the creativity of the original artist’s work while enabling them to earn royalties for their contributions. Moreover, the procurement of licenses safeguards the users of such platforms from any copyright infringement claim securing the stakeholders’ interests.
However, the underlying reason for developing such policies is to avoid venturing into a legal debate over who has the authorship rights to an AI-generated image. As analysed above, the current position of law does not prefer to grant authorship rights to the AI software itself. While the developers of the AI software are a potential candidate to obtain authorship rights, especially in the absence of granting the same to the software itself, it remains contentious due to their lack of contribution to the output. In several cases, either the developer might not have conceived the final output, or the input provided in the form of mere textual prompts might not be significant enough to be acknowledged as a contribution towards the final output. However, this contention can be challenged by providing the developers with authorship rights to incentivise them to create and invest in work. Furthermore, if the law can be settled on providing authorship rights to such developers of software, then Adobe’s model would be become futile and would need to be restructured according to the change in the stance of the law.
The way forward
It is discernible from the above discussion that the conundrum regarding granting of authorship rights needs to be resolved at the earliest by delving into the dilemma from a copyright law perspective. A remedy for the same could be to allow the authorship rights to vest with the developers of the software. The developers of the software exercise high intellectual labour alongside investment in terms of capital and labour to develop such software. It would only be equitable to reward them with authorship rights as a fruit of their intellectual contribution. Alternatively, if the work is put in the public domain, making it accessible to everyone, it would disincentivise the developer to invest or improve the standards of their creation, resulting in diminishing innovation in technology and AI. It would also have a counter-productive effect in terms of reducing the number of works entering the public domain. Therefore, it is argued that the software developer should be provided copyright protection irrespective of their conception of the output as their contribution lies in providing the input, instructions and capability to the AI to produce the output. The inputs provided meet the threshold of originality in the creation of the software and also resolved the issue of not being able to provide rights to the AI software due to a lack of legal personhood by assigning it to a human.
Incorporating such rights of the developer can be done through the “work made for hire” doctrine prevalent in the United States. This doctrine assumes the developer as the employer and the AI as the employee or agent of the employer. Here, the authorship rights are given to the employer who ensures the creation of the work rather than assigning it to the individual who conceived the expression. At face value, the current legal definitions and scheme of things differentiate an employer-employee relationship from the one shared by the developer and the AI due to the lack of both parties being natural persons in the latter case. Nonetheless, with adequate legal reinterpretation to expand the ambit of an employer-employee relationship to be inclusive of non-humans, the same rules which govern an employee can be imposed on the AI. This would also be in furtherance of achieving copyrightable status for AI-generated works and achieving the ultimate goal of granting authorship rights to a human.
Conclusion
Adobe’s policy reflects an acknowledgement of the change in technology and the role played by AI in the modern-day world. While its string of policies emerges as a temporary model, it is a form of escape mechanism as it circumvents the debate regarding granting of authorship rights to images generated by the AI. Although its policy secures its stakeholders’ interests, it also reflects the need to develop laws regarding the authorship of AI-generated imagery in the long run. The existing jurisprudence forbids granting authorship rights to the AI software itself as such rights only vest with a human. In such a scenario, granting authorship rights to the developer of the AI software can resolve the debate. The “work made for hire” doctrine can incorporate such a model into the existing framework. Therefore, the law must reach a settled position, as analysed above, instead of finding models which escape the entire copyright debate.