Abstract :
This research paper deals with the interplay & intersection of AI & IPR as the name suggests. It first talks about the ever-growing role of AI in the creative industry and how AI covers everything from writing assistance to image synthesis. But these props come with their cons. Thus, arises the legal issues & cases of IP rights & copyright infringement which can occur at the level of training data as well as after the AI-generated output is created. It deals with important case laws and precedents because in this scenario these form a major foundation as the existing laws are not fully equipped to resolve these legal conundrums. It encompasses cases like the Andy Warhol Foundation v. Lynn Goldsmith & Stephen Thaler v. Shira Perlmutter which form major precedents in this evolving legal landscape. It also touches upon the legal framework in China & South Korea. It crisply explains the important concepts of fair use & originality. In the end, the research paper takes a look at the current situation from the perspectives of different stakeholders like the AI Platform Developers, Investors, Content Creators, Customers & Businesses using AI Technologies. On top of that, it provides some practical suggestions to solve this legal quandary & highlights the drastic requirements of revamping & updating our legal framework.
Keywords :
AI, IPR, Author, Originality, Fair Use Doctrine, Precedents / Case Laws
Introduction :
Generative AI can seem magical.
AI image generators such as Stable Diffusion, Midjourney & DALL-E can create exceptional visuals in styles. The resulting creations are captivating, both in terms of quality and time for creation are uplifted compared with average human efficiency. The Museum of Modern Art in New York staged an AI-generated creation curated from the museum’s own collection.
The capacity of AI-powered writing assistants is even more profound, as they write essays, articles, summaries, poems, music and novels and are proving to be proficient in emulation of style and form.
Now even if it may seem that these new AI tools can incant new creations from out of thin air, that’s not the case. Generative AI platforms are trained using data lakes and question snippets. Billions of parameters are developed by the software which processes huge repositories of images and text. The AI platforms recover patterns and relationships, which they later utilize to form rules and make judgments and predictions when responding to a prompt.
Generative AI is thus becoming more widespread in the creative, academic & business fields.
However, the legal implications of using generative AI are still ambiguous. This process involves legal risks particularly –
- Intellectual property infringement which includes copyright, patent & trademark infringement. This infringement could be at two levels :
- At the primary level using unlicensed content as training data for AI models.
- At the secondary level by using unauthorized work as a base and then creating derivative works from it without significant transformative power.
- Another major issue is regarding the ownership of AI-generated works & whether IP rights could be granted to such works.
In many cases, it also poses legal questions that are still being resolved.
Research Methodology :
This research paper explores as the name suggests the interplay between Artificial Intelligence (AI) and Intellectual Property Rights (IPR) using a multi-faceted approach.
To explore this avenue this research employed the method of content analysis.
To gain a comprehensive understanding of this evolving landscape, I have analysed various digital articles offering current perspectives and industry insights. Additionally, I delved into case laws and legal precedents to understand how courts have grappled with AI-related IPR issues. Finally, to ground the research in the current legal framework, I examined relevant statutes that govern intellectual property. The research further considered both opposing and supporting viewpoints on key topics within this domain. Not to forget that my initial interest in this topic was sparked by a recent national conference on the very same subject, which provided a valuable foundation for further exploration & my entire research.
Literature Review :
In the digital article entitled Generative AI Has an Intellectual Property Problem by Gil Appel, Juliana Neelbauer and David A. Schweidel, the authors begin by unpacking the inner workings of Generative AI tools, detailing both their making (training data) and function. They then delve into the legal complexities that arise from these very functionalities & its making, exploring how the essence of these AI tools themselves sparks a multitude of legal questions. The article discusses the importance of the doctrine of fair use and other cornerstone legal concepts, including the degree of transformative use, the question of originality, and the nature of derivative works in this context. Furthermore, it adopts a multifaceted approach, examining this issue from the perspectives of various stakeholders – the developers of these tools & the customers or the very users who interact with Generative AI. Importantly, the article acknowledges the anxieties of creative artists and authors regarding the protection of their intellectual property rights. Finally, the authors propose solutions to navigate this intricate situation.
The WIPO’s official website also proved to be of great help during this research on Interplay between AI & IPR. The website offers an immense amount of intellectual & authentic data. It also includes a lot of scholarly articles which give a deep insight into the respective legal conundrums. I have referred to a few of these articles during my research.
Case Laws & the evolving landscape of IPR :
The above aspects & claims are already being litigated with the courts solving the puzzle of how well-established laws should be employed in this current baffling situation.
Andersen v. Stability AI Ltd., et al. (2022)
Here three artists (Sarah Andersen, Kelly McKernan and Karla Ortiz) filed a putative class action lawsuit to sue multiple Generative AI platforms (Stability AI Ltd., Stability AI Inc., Midjourney Inc., and DeviantArt Inc.) alleging copyright infringement and violation of the Digital Millennium Copyright Act (DMCA).
The fundamental assertion centred around the accusation that the defendants’ AI models were trained on data lakes containing millions of copyrighted images, including works by the plaintiffs, without permission.
The plaintiffs argued that the AI-generated outputs derived from these copyrighted works instituted infringement and that the training procedure itself violated their copyright.
In this case, the court’s decision to allow the plaintiff’s direct copyright infringement claim sets a precedent for future cases in this evolving field of law.
Getty Images v. Stability AI. (2023)
Getty Images an American visual media company alleges Stability AI of infringing its Intellectual Property Rights in two ways :
Getty claims Stability AI “scraped” millions of images and metadata from Getty’s website without permission to train Stable Diffusion through its data lakes.
Getty argues that the then AI-generated images by Stable Diffusion reproduce substantial parts of their copyrighted works and even incorporate Getty’s watermarks, constituting trademark infringement.
The case is in progress & during its initial hearing, the court rejected Stability AI’s motion to dismiss the lawsuit.
The court ordered that Getty’s copyright claims can proceed to trial.
Other risks :
There lies an increased risk of unconscious sharing of confidential trade secrets and business information by putting data into data lakes of Generative AI platforms.
IP Rights :
In such a situation, arises the question of the moral rights like the right to attribution and the right to integrity granted to the authors & artists under the IP rights.
The right to attribution is the claim of authorship of their work & it ensures that the creators receive public recognition for their work.
The right to integrity protects the moral integrity of a creator’s work. It allows the creators to object to any distortion, mutilation, or other modification of their work that could harm their reputation or the work’s meaning.
The Doctrine of Fair Use :
The result of all such cases is dependent on the interpretation of the fair use doctrine. This allows copyrighted work to be utilized without the owner’s permission “for purposes such as criticism, comment, news reporting, teaching, scholarship, or research,” and for transformative use which allows the copyrighted material to be used for a new purpose and meaning, creating something original rather than a mere copy or substitute.
In the context of the United States, The Copyright Act of 1976 codifies the fair use doctrine, which was evolved by the courts. When deciding whether or not a particular work falls under the Act’s “fair use” provisions, four criteria must be considered.
They are as follows:
▪ The Purpose and Character of the Use.
▪ The Nature of the Copyrighted Work.
▪ The Amount or Substantiality of the Portion Used.
▪ The Effect of the Use on the Potential Market for or Value of the Work.
In the Indian context, The Indian Copyright Act of 1957 (ICA) allows certain copyrighted material to be used without the permission of the copyright holder. The principles of fair use doctrine are mentioned in Section 52 of the ICA. The “Doctrine of fair use” evolved as courts balanced between the IPR and the public’s interest in permitting copying in certain genuine circumstances.
Fair use does not have any specific definition instead it is a very subjective concept and can vary in every other case. The main motive back of this concept is that not all copying be banned, for example: for a socially impactful purpose.
Authors Guild v. Google, Inc. (2015)
This was a landmark case centred around the legality of Google’s Library Project to form an online database called Google Books. The Authors Guild, representing authors and publishers, sued Google for copyright infringement, alleging that Google’s large-scale book scanning and displaying snippets of text from copyrighted books was a massive unauthorized use of copyrighted material.
But Google stated that their utilization was protected by the principle of fair use, as it was just collecting copyrighted works to put on an online searchable library for research, education & other scholarly purposes.
The court decided this case on the basis of four criteria’s of fair use from Section 107 of the U.S. Copyright Act:
On the basis of this analysis, the court ruled in favor of Google, stating that Google’s use of copyrighted material was protected by fair use.
Moreover, Google only displayed snippets & parts of copyrighted material and provided limited access to full text.
But now this is something contrasting to our current case related to AI & IPR, as here the developers of AI Platforms use a substantial portion of copyrighted material.
But there are other, non-technological cases that could shape how the products of generative AI are treated.
Andy Warhol Foundation vs. Lynn Goldsmith (2023)
This case before the U.S. Supreme Court against the Andy Warhol Foundation (AWF) was brought by the photographer Lynn Goldsmith. AWF created a series of silk screen images based on a photograph of Musician Prince taken by Lynn Goldsmith. Goldsmith claimed Warhol’s work infringed her copyright over the original photograph. However, AWF argued that it was a fair use being a transformative work.
The court ruled in favour of Lynn Goldsmith, stating that Warhol’s work did not fulfil the bare minimum of transformation for taking the defence of fair use. It included the very essence & the most essential element of Goldsmith’s Photo
The case thus lifted the level of transformative use & emphasized that simply changing the style or aesthetics of a copyrighted work is not enough.
Legal Framework in China :
Li Yunkai v. Liu Yuanchun
The plaintiff created images through prompts in an AI Text-to-image software and shared them on his social media account. Later, the defendant used these on a different social media platform, removing the plaintiff’s watermark. Thus, the plaintiff claimed infringement of his IP Rights, leading to legal proceedings.
The plaintiff actively created elements, layouts & compositions of the images via his prompts & followed an iterative process which reflects his personal choices showcasing originality.
Thus, the court gave authorship & copyright to the plaintiff (AI tool user) & considered AI as a mere tool for human creativity.
South Korea’s outlook :
Recently, South Korea’s Ministry of Culture, Sports and Tourism, responsible for IPR policy, announced the publishing of “AI Copyright Guidebook & said that AI-generated work will not be given copyright unless it “demonstrably conveys human thoughts and emotion”.
Copyright infringement by AI-generated output :
Stephen Thaler v. Shira Perlmutter, Register of Copyrights & Director of United States Copyright Office, et al.
Mr. Stephen Thaler applied for the copyright of “A Recent Entrance to Paradise“, a visual art autonomously generated by his AI system. But the Copyright Office rejected his application, for the work lacked human authorship. Thaler countered this decision many times, but the Copyright Office stood firm.
Thaler then filed a lawsuit challenging this decision.
The United States District Court for the District of Columbia examined whether an autonomously generated AI work is copyrightable under the Copyright Act of 1976.
With an order on the 18th of August 2023, the court affirmed the Copyright Office’s rejection of application for the copyright registration & stated that “the copyright law’s bedrock requirement is human authorship citing centuries of settled understanding.”
The act protects original works of authorship & defines it as “fixed in a tangible medium of expression when its embodiment in a copy or phonorecord, by or under the authority of the author, is sufficiently permanent or stable to permit it to be perceived, reproduced, or otherwise communicated for a period of more than transitory duration.” Here the “fixing” of the work must be done “by or under the authority of the author”. Thus, the crucial requirement is that the work must have an “author”.
Encouraging creativity in people & promoting science and arts, was the main aim of American copyright laws from its inception, as non-human actors need no incentivisation or IP Rights.
The court concluded that copyright protection will not be granted to works with no human involvement. Thus, the plaintiff’s autonomously generated work was not granted copyright.
UK’s Stance :
UK’s Copyright Designs and Patents Act 1988 gives copyright protection to computer-generated works with no human author. The law says that a work “generated by computer in circumstances where there is no human author“, the author in such a case is “the person by whom the arrangements necessary for the creation of the work are undertaken”.
This entire issue regarding the copyrightability of AI-generated content revolves around interpreting the term author or artist & originality.
Now even if we grant copyright to AI-generated content, we are faced with a brand-new question of who the law should consider to be the person/author responsible for the work to be generated. Would it be the AI programmer/developer or the user of that program giving the prompts?
Here while discussing IPR & Copyright we cannot forget the Berne Convention for the Protection of Literary and Artistic Works. This is an international treaty that stands as the cornerstone in the arena of copyright protection across member countries.
Given this situation of legal quandary, we require a radical altering of IPR Laws within the framework of the Berne Convention
Suggestions :
A recent effort to engage & interact on IP policy aspects of AI, the WIPO Conversation on IP and AI commenced in September 2019 to discuss the impact of AI on IP policy.
Measures from the viewpoints of different stakeholders in today’s digital economy concerning the intersection of AI & IPR.
AI Developers :
Stable Diffusion developed by Stability AI, declared that “artists will be able to opt-out in the next generation of the AI image Generator.”
But this decision has two major flaws :
- The AI developers are evading their responsibility to license the IP before using it & are putting the onus on the content creators to protect it. Instead, the companies should give the creators an option to opt-in rather than opt-out.
- This particular decision to opt-out applies only to the next generation of the AI platform
Thus, AI developers must make sure they follow the law while acquiring data utilized to train their AI by licensing and compensating the copyright holders.
Developers should maintain the origin of AI-generated content by recording the source of the training data and tracking the seed data’s metadata to level up the training data’s transparency.
Insurance companies & Investors :
In future, insurance companies to grant business insurance coverage & investors to invest may demand transparency reports if the company’s assets consist of AI-generated works.
Creators :
Creators need to actively search for their copyrighted work in the AI Platforms data lakes, in fact some technologies can even help in obscurement from these algorithms.
Content creators must check social channels for works similar to their which could have been derived from their own.
Trademark Infringement :
Brands must not only look for their specific trademark elements like the Nike Swoosh or the iconic castle logo of Disney because even if the specific elements like the logo itself might not be present, the other stylistic elements which indicate the brand’s identity might be used. And this shows the intent to deceive customers by mimicking a famous brand’s trademark elements.
Customers :
Users of these AI tools must enquire to the developers about the kind of data used for training their AI tools and review the terms & conditions & the platform’s privacy policy.
Users must also avoid using AI platforms which are not transparent enough about the licensing of their training data from the respective content creators or are using unlicenced data instead of using the open-source licensed data which the AI companies comply with or claim to do so.
Businesses :
There are multiple benefits of generative AI but at the same time, businesses need to consider the potential risks of utilizing AI-generated content in their workplaces & know how to protect themselves from such risks.
As an initial step, businesses need to ask for the terms & conditions of the services of the AI platforms to ensure that the platform’s training data is properly licensed & is authentic.
Also, they should demand adequate compensation in case if there occurs IPR infringement in the future caused by the usage of the AI Platform due to the failure of the platform in properly licensing its training data from the copyright holders.
Conclusion :
The entire frontier of AI creativity is full of opportunities, and navigating it seems like finding an equilibrium between rights and incentivization. The legal issues around ‘AI-made stuff’ are completely baffling, especially when it comes to who gets the ownership & IP Rights. Traditional laws just weren’t built to deal with these kinds of questions. The bottom line is that the whole authorship and originality concept will play a major role in being the decision maker for granting AI creations copyright protection. This whole paradigm shift needs us to look again & reevaluate the existing laws consciously. We need a revamped legal system that lets innovation flourish while also safeguarding intellectual property rights. In a nutshell, a progressive-thinking legal framework is key to unlocking the full potential of AI. This isn’t just about updating laws for new tech, it’s about building a dynamic and inclusive legal ecosystem, both here in India and worldwide.
Name: Kanishka Khamkar
College Name: ILS Law College, Pune.