The interaction between intellectual property laws and AI: Opportunities and challenges

Introduction

In recent years, rapid developments in AI systems and technologies have transformed AI from a futuristic concept into an integral part of business strategy. Companies across various industries are increasingly eager to incorporate AI into their operations, driven by the promise of enhanced efficiency and improved decision-making capabilities. Global spending on AI-related applications, services, and infrastructure are at an all-time high, with IDC predicting that spending will more than double by 2028 to USD 632 billion, with a compound annual growth rate of 40.6 percent from 2023 to 2028¹. 

Generative AI, or “GenAI” for short, is a vital component of the AI market. GenAI systems are trained to recognise patterns and relationships within data using staggeringly large and diverse data sets; this allows GenAI systems to produce new content that can pass off as human-generated. With GenAI applications such as ChatGPT now widely available at low or no cost, the average person can easily create new content that in previous years would have taken a significant amount of time to produce. According to a recent McKinsey global survey, as of May 2024, 65 percent of organisations are regularly using GenAI in at least one business function, nearly double the percentage from their previous survey in 2023². Ernst & Young estimates that a GenAI-driven productivity upswing could result in global GDP increasing by almost USD 1.2 trillion to USD 2.4 trillion over the next decade.³

There is no doubt that AI systems are here to stay, and that the use of AI will fundamentally reshape business operations in the years to come. However, the development and use of AI technologies have generated much controversy as well, especially in the field of intellectual property where numerous high profile AI-related cases have been fought in court.

Many countries across the word have grappled – and continue to grapple – with several key issues, including:

• Can the content created by an AI system be protected under IP laws? 
• Who owns the IP rights in a piece of work that has been created by an AI system?
• Is there a risk of IP infringement risk when using an GenAI system?

In this article, we will analyse these issues and explore the potential implications for human inventors, authors, right holders and the IP legal system.

IP protection and AI

IP laws grant IP owners certain rights to exclude others. For instance, patent owners can prevent others from using their inventions without permission, and copyright owners can prevent others from copying their works without permission. These rights enable IP owners to be recognised and derive financial benefit from their creations,⁴ providing them with an incentive to invest their time, effort and resources into their intellectual creations.

Traditionally, this has been one of the key justifications for IP laws. However, this reasoning does not apply to AI systems, which do not require an incentive to create.⁵ It is therefore apt to ask whether IP laws should apply to the creations of AI systems, in the same way that they apply to works created by human hands. Our discussion will focus on patent and copyright laws, these being the two areas of IP law that are most affected by the AI revolution.

A. The Requirement of Human Involvement

Patent laws generally require an inventor to be named when filing a patent application. Without a named inventor, a patent cannot be issued and the application will fail.

Can an AI system be recognised as an inventor of a patent? This was put to the test when Dr Stephen Thaler filed patent applications across multiple countries designating DABUS, an AI system owned and created by him, as the sole inventor.⁶ Almost all of these applications have been rejected on the basis that patent laws – following the plain meaning of the words used in these laws – require a natural human to be named as an inventor. For instance, the European Patent Office concluded that the inventor of a patent must be a human because the “family name, given names and full addresses” of the inventor must be named.⁷ The general consensus, therefore, is that an AI system cannot be named as a patent inventor.

The position is similar under copyright law, which protects original works of authorship, such as literary, artistic, musical, and dramatic works. Unlike patent law, copyright law does not require an author to be named for a work to enjoy copyright protection. However, most copyright laws imply or assume that an author must be a natural person. For instance, the copyright in authorial works typically lasts for the lifetime of the author plus a specified number of years.⁸ This cannot be applicable to AI systems which are not living beings; otherwise, works produced by AI systems would enjoy perpetual copyright protection. 

While an AI system cannot be the inventor of a patent or the author of a copyright work, this does not mean that inventions and works created by AI systems can never be protected under IP laws. In this regard, it is helpful to distinguish between AI-assisted content, where AI is merely used as a tool or assistant which helps a human to create an invention or work, and AI-generated content, where an invention or work has been generated by an AI system with limited or no human input.9

Conceptually, it is easier to obtain IP protection over AI-assisted content, as there is obvious human involvement in the creation of the content.¹⁰ Where a human inventor or author uses AI to assist them in the creative process, this is arguably no different from using any other electronic tool; the use of AI in this case should not be a barrier to obtaining IP protection.

However, the issue of whether AI-generated content should be protected by IP laws is more controversial, given that the level of human involvement in creating the content is less (and in some situations, minimal). The answer will depend heavily on the factual matrix, and approaches will diverge across different countries.

By way of illustration, in the U.S., which allows registration of copyright works, the U.S. Copyright Office rejected the registration of artist Jason Allen’s piece titled “Théâtre D’opéra Spatial,” which was generated using Midjourney, a text-to-image GenAI tool. This was despite Mr Allen claiming to have inputted “at least 624” text prompts to adjust the scene, tone and focus of the image, and using another platform, Gigapixel AI, to increase its resolution and size.¹¹ In China, however, the Beijing Internet Court held that copyright subsisted in a piece of artwork generated through the text-to-image software Stable Diffusion, due to the plaintiff’s “intellectual investment” in designing the presentation of the character, selecting and arranging the order of prompt words, setting parameters and selecting the picture that he wanted.¹²

B. Who is the IP Owner?

In addition to the tricky issue of whether an AI-generated creation can be protected by IP laws, another hotly debated (and closely related) issue is who should own the IP rights in AI-generated content. This is more than an academic debate. IP ownership is significant, because the owner of an IP right is entitled to permit or exclude others from using that right, and has the legal standing to sue if that right is infringed.

Most stakeholders agree that AI systems do not have legal capacity or personality, and are therefore incapable of owning IP rights. Beyond that, there is a diverse range of opinions on who the IP rights owner should be – or if there should be a rights owner at all.

Possible candidates for ownership of an AI-generated invention or work include the owner of the AI system; the system developer who has coded the AI system; the trainer of the AI system; the user or operator of the AI system; or some permutation of the above. It is also possible to assign IP ownership through contractual provisions – for instance, the terms of use for OpenAI’s services, which include ChatGPT and DALL-E, expressly assign the IP rights to any AI-generated content to the user: 

“As between you and OpenAI, and to the extent permitted by applicable law, you (a) retain your ownership rights in Input and (b) own the Output. We hereby assign to you all our right, title, and interest, if any, in and to Output.”¹³

Another point of view is that AI-generated content should be in the public domain.¹⁴ The United States Copyright Office has published guidelines on works containing AI-generated material,¹⁵ stating that “When an AI technology determines the expressive elements of its output, the generated material is not the product of human authorship. As a result, that material is not protected by copyright and must be disclaimed in a registration application.” Based on these guidelines, where an AI system produces text or image solely based on a prompt from a human, the output will not be protected by copyright.¹⁶

At a national level, the best way to provide legal certainty over these thorny issues may be through legislation. Notably, some countries have chosen to recognise sui generis rights in AI-generated content. For instance, Section 9(3) of the UK Copyright, Designs and Patents Act 1988 provides that in the case of a literary, dramatic, musical or artistic work which is computer-generated, the author (and the first owner of the copyright in the work) is “the person by whom the arrangements necessary for the creation of the work are undertaken.”¹⁷ Ukraine has gone arguably further, having amended its copyright law in December 2022 to grant special rights over non-original output created by AI systems.  These rights either belong to the individual who has the software licence that created the work or the software owner.¹⁸

IP infringement risks of using AI

When using AI technology to generate content, another important consideration is whether there is a risk of infringing third party IP rights. The risk of infringement cannot be overstated. In recent years, there have been numerous high-profile IP infringement cases involving the use of AI systems – examples include Getty Images v Stability AI and Zhang v Google.

While these cases have focused on the provider or owner of the AI system, perhaps because they have the deepest pockets, the risk of IP infringement is present at other levels of the AI supply chain and affects multiple stakeholders. We examine those risks in greater detail below.

A. Training AI Systems

GenAI systems are trained using massive amounts of data, much of which is publicly available on the internet. If any of this data is protected by copyright or other IP rights, the use of such data to train a GenAI system would infringe these rights where carried out without the permission of the rights owners.

While certain countries have implemented exceptions to IP infringement which may be relevant when training a GenAI system,¹⁹ many countries have not. For GenAI systems that are trained on a global pool of data, these exceptions are therefore unlikely to serve as a get-out-of-jail-free card, since the risk of infringement will always be present in some countries.

B. Creating and Using Output Generated by AI Systems

Infringement risks are also inherent in the process of creating and using output generated by AI systems. From a copyright law perspective, where a person creates content that is identical or substantially similar to a copyright work, this is typically taken as evidence of copying. Where an AI system can only rely on a small number of data sources to generate content – for instance, when asked to respond to a particularly niche question – it is possible that the content generated will replicate large parts of those data sources, or even be identical to those sources. 

C. Who is Potentially Liable?

While the AI system itself lacks legal personality and cannot be held liable for infringement, the stakeholders involved in training the AI system, or creating and using its output, will be at risk. These stakeholders may include:

• the developer / programmer of the AI system, who developed the code and algorithms on which the AI system trains and operates;
• the owner of the AI system, who may be held vicariously liable for the acts of the AI system;
• a corporate entity that deploys the AI system (and may possibly develop and train the AI system in its own IT environment); and
• the individual end users of the AI system, who would use the infringing system to create content, and subsequently use that content for their own purposes.²⁰

This does not mean that each of the stakeholders will bear the same level of infringement risk. The legal position will also vary significantly across different countries. However, all entities involved in the AI ecosystem would do well to be aware of the potential risks.

The laws governing IP infringement for using AI systems, and who is liable for infringement, will evolve rapidly as more cases are litigated and countries provide guidance (or implement new laws) to address controversial issues. In the meantime, it would be prudent to consider taking measures to mitigate the risks of IP infringement when using AI systems. These include:

• for providers and deployers of AI systems, fine-tuning the systems to minimise the risk that identical copies of existing data sources will be delivered as output.
• for users of AI-generated content, using plagiarism-checking software to identify if the content is materially similar to copyright works owned by third parties.
• training employees to avoid risky practices, such as asking an AI system to copy a competitor’s marketing literature or product description.
• for corporate entities that intend to deploy AI in their organisations, seeking contractual protection (e.g. warranties or indemnities) from the AI system provider / owner against third party infringement claims.²¹

Conclusion

As AI technologies continue to develop, they will create new opportunities and challenges for IP protection and enforcement. The legal landscape will change as courts and regulators grapple with the complex and novel issues outlined above. As such, all stakeholders in the AI ecosystem should keep abreast of the latest legal developments and trends, and seek expert advice if they are unsure of how to navigate the legal risks.