Intellectual property rights: a barrier to innovation in GenAI?

June 2024 | SPOTLIGHT | INTELLECTUAL PROPERTY

Financier Worldwide Magazine

June 2024 Issue

In the past two years, the proliferation of text to image and large language models available for public use has led to a growing interest among businesses in integrating generative AI (GenAI) into their operations.

According to Statista, the value of the GenAI industry is estimated to reach $36bn in 2024, with a projected annual growth rate of 46 percent. Investments in GenAI software, services and hardware have reached record highs, with the International Data Corporation predicting that the Asia-Pacific region alone will spend $26bn by 2027.

A significant part of GenAI’s appeal is its ability to create new ideas and products in a fraction of the time that a human would take. AI systems, such as generative adversarial networks and generative pre-trained transformers, are trained using large data sets, allowing them to produce output that can pass off as human-generated.

There are potential GenAI use cases across a wide range of industries. In the pharmaceutical industry, GenAI has discovered and developed molecules that are already in clinical trials. A recent Gartner report found that 55 percent of organisations are either piloting or implementing GenAI solutions in their business.

GenAI, it would seem, is poised to disrupt industries and substantively reshape business operations around the world. However, amid this surge in popularity, it is important to ask the fundamental question of how content produced by GenAI will be protected under intellectual property (IP) laws.

IP laws grant certain rights to IP owners; for instance, patent owners are entitled to prevent others from using their inventions without permission, and copyright owners are entitled to prevent others from copying their works without permission. Without IP protection, the value of an invention or a copyrighted work would be greatly diminished, because there is nothing to stop third parties from replicating the invention or work.

Traditionally, patent and copyright laws have required the involvement of human creators before IP rights can subsist in an invention or work. Where machines such as GenAI systems autonomously generate an invention or work, will these be protected under IP laws?

Redefining innovation: can a GenAI system create a patentable invention?

Patent laws generally require an inventor to be named as part of a patent application. If a GenAI system cannot be named as an inventor in patent applications, it raises the issue of whether an invention created by a GenAI system can be protected under patent law at all, even if it meets all other conditions of patentability.

This issue was considered by various IP offices and courts over the past five years when Dr Stephen Thaler filed patent applications in multiple countries designating ‘DABUS’, an AI system he owned and created, as the sole inventor. Most of these applications have been rejected on the basis that patent laws only allow a human to be named as an inventor. This is because the words used in these patent laws imply that the inventor must be a human being.

For instance, courts have pointed to the pronouns used in these laws to refer to the inventor, indicating that the inventor must be a natural person. The European Patent Office concluded that in order for an invention to qualify for patent protection, the inventor must be a human because the family name, given names and full addresses of the inventor must be named.

The general consensus, therefore, is that GenAI systems cannot be named as patent inventors. As GenAI systems do not have legal capacity or personality, they cannot be patent owners either. However, there is considerable debate on whether the owner, developer or operator of a GenAI system could be the owner of an invention created using GenAI.

There is a distinction between ownership and inventorship – while inventorship gives a person a right to be named in the patent, ownership allows a patent owner to decide who may or may not use the patented invention, and to sue for infringement of the patent. In countries where the position is unclear, the only way to settle this debate may ultimately be through legal reform.

Redefining creativity: can a GenAI system create a copyrighted work?

Just as GenAI systems cannot be named as patent inventors, they also cannot be named as authors of copyrighted works under most existing copyright laws.

Similar to patent laws, the text of most copyright laws assumes or implies that an author must be a natural person. Copyright in authorial works typically subsists for the lifetime of the author plus a fixed number of years; as machines cannot die, allowing a GenAI system to be an author would grant perpetual copyright protection to its works.

Furthermore, copyright is only granted to works that are original. Originality – which means that a work was independently created – implies some level of human creativity. Indeed, a fundamental principle of the copyright system is to reward and incentivise the expression of such creativity. Without a human author, it becomes challenging to establish the necessary originality and authorship required for copyright protection.

Does this mean that works created by GenAI systems can never enjoy copyright protection? Not quite. Depending on how the GenAI system in question operates and the level of human input provided (e.g., by way of prompts), it may be argued that the GenAI system is merely a tool or assistant which helps a human to create a work. In this scenario, the human – who could be the operator or user of the GenAI system – will be the author of the work.

Whether such arguments are successful will be heavily dependent on the facts, and views may differ significantly across countries. In the US, where copyrighted works can be registered, the US Copyright Office rejected the registration of artist Jason Allen’s piece titled ‘Théâtre D’opéra Spatial’, which used Midjourney, a text to image GenAI tool.

This was despite the artist’s claim of inputting 624 text prompts and using Adobe Photoshop to craft new visual elements. In China, however, the Beijing court in Li Yunkai v. Liu Yuanchun held that copyright subsisted in a piece of artwork generated through Stable Diffusion, due to the plaintiff’s “intellectual investment” in designing the presentation of the character, selecting and arranging the order of prompt words (including 120 negative prompts), and choosing parameters.

Redefining legal frameworks: the way forward?

Traditional patent and copyright law frameworks have been in place well before the advent and proliferation of GenAI systems. They were designed to protect the rights of human inventors and creators, and it is not surprising that inventions and works produced autonomously by a GenAI system fit awkwardly within these frameworks.

If it is not clear how AI-generated inventions and works will be protected under IP laws, AI developers and deployers may be deterred from further improving GenAI technology. Building a GenAI system is costly and resource intensive. If AI-generated inventions and works do not enjoy IP protection, there may be less incentive to invest time, resources and expertise into developing and refining GenAI systems, ultimately stifling innovation in this rapidly growing industry.

For businesses that use GenAI to create and develop work products, if there are no IP rights in these work products, they will not be able to prevent third parties from copying or replicating them. This may make some businesses hesitant to incorporate GenAI into their work processes, even where the use of GenAI will result in a significant improvement in efficiency.

Addressing these issues is not easy. Existing IP laws and frameworks may have to be adjusted, or special exceptions made, to recognise that AI-generated works and inventions should be treated differently from conventional works and inventions.

The good news is that the international community is having frank discussions on these issues. The World Intellectual Property Organization is leading the way in these discussions, exploring questions such as whether AI innovation and creation require IP incentives, and how to strike a balance between the value of human invention and creation and AI innovation and creation.

In time to come, changes to legal frameworks and policies may take place on a widespread national level. In the meantime, stakeholders in the GenAI industry, including users, developers and operators, would do well to be aware of the possible complexities when claiming IP protection over AI-generated works.

Justin Davidson is a partner at Norton Rose Fulbright and Jeremiah Chew is a director at Ascendant Legal LLC. Mr Davidson can be contacted on +852 3405 2426 or by email: justin.davidson@nortonrosefulbright.com. Mr Chew can be contacted on +65 6309 5414 or by email: jeremiah.chew@nortonrosefulbright.com.

Justin Davidson

Norton Rose Fulbright

Jeremiah Chew

Ascendant Legal LLC