Arushi Dubey[1]
1. Introduction
Imagine a scenario where a novice without expertise in music composing uses Artificial Intelligence (AI) tools to create music, uploads the AI-generated musical tune on social media platforms and the tune becomes viral overnight. Who in this scenario will be the owner of that music: the novice who gave an appropriate command to the AI tool, the AI which created the music, or the software developer who created the AI? How will the copyright owners of the underlying works that were used to train the AI tool be compensated? It is pertinent to answer these questions. from a copyright law point of view.
The musical plot envisioned in the beginning has been made possible for different types of creative works from literary to artistic by AI tools like ChatGPT and DALL-E. The artificial intelligence company owned by Google, DeepMind, has created a software called Wave Net which can create music by listening to recordings. This software is not yet publicly available, but with Google launching Bard as its answer to ChatGPT, it is only a matter of time before Google will also release the other AI tools that it is developing to the public.
2. Copyright law and artificial intelligence: a sticky saga
Copyright law is one of the areas in which AI technology like ChatGPT and DALL-E interact, which poses several legal questions. The first tryst of copyright law and AI occurred when Dr. Stephen Thaler sought copyright registration for a piece of art called ‘The Recent Entrance to Paradise’ created by the AI algorithm Creativity Machine. Dr. Thaler filed an application to register the artistic creation with the US Copyright Office. In the application he stated that Creativity Machine was the author, and he was the claimant. The application was accompanied with a transfer statement stating that he had acquired the ownership of the work as he was the owner of the machine. However, his application was rejected because under Sections 202.02(b) and 503.03(a) of the Compendium II of Copyright Office Practices, a machine cannot be the author in the USA. This led to Dr. Thaler asking the U.S. District Court to overturn the rejection of his copyright application.[2]
In the UK, copyright law offers a first of its kind protection to computer generated works created without human interference. Section 9(3) of the Copyright, Design, Patent Act, 1988 (CDPA) states that the author of these works will be the ‘person by whom the arrangements necessary for the creation of the works are undertaken’. However, this unique provision is not without its problems. The requirement of originality in copyright law is what makes the road to copyright protection for AI-generated works difficult. As per Section 1(1)(a) of the CDPA, for any literary, dramatic, musical, and artistic work to qualify for copyright protection, it must be original. The UK has followed the traditional approach to test originality. Which states that for any work to be original, it should not have been copied and it should have been created by the skill, labour, and judgement of the author.
On the other hand , the European Union’s approach of testing originality is on the basis of the work being author’s own intellectual creation.[3] In the seminal case of Temple Island Collections Ltd v New English Teas Ltd ([2012] EWPCC 1), both of these concepts were used interchangeably by the English court.[4] The test of originality in this case was based on human authorship in the work and the investment of the human’s skill, judgement and intellectual prowess in creating it. It is this personal touch which can be found lacking in computer-generated works, which makes this a grey area of law.[5]
3. The Ownership -Authorship Question
The main impediment to protecting AI-generated works under copyright law across jurisdictions, is that copyright is comfortably only granted to original works made by humans. Even when a computer-generated work is given protection under copyright law, the author/owner of the works is the person who causes it to be created. Meaning thereby that AI would not be given the copyright for AI-generated content. The developer of the AI application would also not get the copyright since he/she created the AI, not the content. Even the person giving the command to the AI application to produce the content would not get copyright as per law. As a mere command by any person to an AI-generative platform would be not be considered author’s own intellectual creation nor will it be considered as being produced by the person’s skill or labour – thus falling short of fulfilling even the bare minimum for obtaining copyright protection. In such a scenario a provision on the lines of Section 9(3) CDPA will not be able to provide a solution. Application of this section would firstly require a determination as to who made ‘the arrangements necessary for the creation of the works’, which in the present situation of public large-scale use of AI tools will be difficult to decide. A possible respite to this impasse has been presented by Dr. Thaler before the US court whereby a creation made by AI can be caught by copyright law under certain legal theories, like the accession doctrine and work-made-for-hire doctrine. This case has the potential to decide the fate of AI generated works across international jurisdictions.
An affirmative decision in favour of Dr Thaler in this case would open the pandora’s box. Which will put impetus on the law makers worldwide to legislate on the question that if an AI work, which has been produced by inputs from more than one person, is copyrightable, who owns the copyright? According to the author, giving copyright control to content generated through AI to only one person would lead to loss of incentive for other players involved in the AI space. Thus, distinct rights and their value would have to be specified in copyright law where different players are involved in creating content through AI. This is not unprecedented; for example, in the music industry, different players are involved in producing content and their efforts are individually acknowledged and rewarded by copyright law across jurisdictions. This multi-right approach will not only help in keeping copyright law relevant in these changing times, but it will also incentivise inventions and content generation in the AI space.
4. The Infringement Question
The ownership of AI creations is not the only concern; infringement is also a big problem. Various cases have been filed on this question across international jurisdictions. In the UK, Getty Images filed a copyright infringement case against Stability AI, alleging that Stability AI used millions of copyright-protected images to train its AI tool Stable Diffusion without proper authorisation. Similarly, a case (Andersen v. Stability AI Ltd., No. 3:23-CV-00201 N.D. Cal. Jan 13, 2023) has been filed in a California federal court by a trio of independent artists alleging that DeviantArt, Midjourney and Stability AI used their art without authorisation to train AI image-generating models and produce ‘new art’.
The question of infringement stems from the way these AI-content-generating algorithms work. It is by analysing and scrapping data on the Internet through which AI is trained and it generates outputs. The data which is used in this process could be protected by copyright law. Which means that the resultant AI output/work is not original, rather it is infringing upon someone else’s works. AI developers argue that this should not be a concern because the nature of the creations/output generated by AI is transformative in nature making it fall under the fair dealing/ fair use defence. Moreover, any similarity between AI creations and any other works would be too small since this technology bases its output on analysing millions of works. Most probably, US courts will follow the decision in Authors Guild Inc v Google Inc (804 F.3d 202 (2d Cir. 2015), which decided that scrapping data available on the internet where it can be viewed openly by the public is not copyright infringement. However, the involvement of AI technology in the present cases complicates matters. These cases will be closely watched by the technological and the legal community.
The pioneering streak established by the UK when it gave copyright protection to computer generated works was maintained when it came to detangling the complicated legal situation created by AI generated works. The UK Intellectual Property Office through its consultation paper on Artificial Intelligence and Intellectual Property: copyright and patents anticipated these legal issues and proposed to bring in a new copyright exception for text and data mining purposes, which would allow developers to train AI models without the threat of being slapped with an infringement lawsuit. However, these proposed changes did not paint the same positive picture when seen through the lens of the likes of authors, publishers, and musicians. In fact, the UK music industry labelled this proposal as ‘dangerous and damaging’because it would prioritise the interest of AI developers over content creators. Keeping these criticisms in mind d DCMS minister announced that the proposal will not be taken forward by the UK government. The music industry has welcomed the announcement stating that their efforts have brought fruit in stalling ‘music laundering’. The consequences that stalling of this proposal will have on the AI innovation landscape are yet to be seen.
5. Conclusion
When it comes to technological developments, the sky is the limit. AI has the possibility to refashion the way humans make and consume content. Strong copyright law which balances the rights of the parties involved in this issue, is the need of the hour. This will provide assurance and clarity to all parties involved, from the people, who rely on making, performing, and distributing content for their bread and butter to those, who are invested in the field of AI.
[1] Advocate, India; King’s College London alumna.
[2] Thaler v. U.S. Copyright Office, No.1:22-CV-01564-BAH, 2023 WL 2889578 (D.D.C. Feb 7, 2023).
[3] Lionel Bently and others, Intellectual Property Law: I Copy Therefore I Am (5th edn, Oxford University Press 2021) 93.
[4]Tanya Aplin and Jennifer Davis, Intellectual Property Law: Texts, Cases and Materials (4th edn, Oxford University Press 2021) 125.
[5] Bently (n 3) 117.