Are AI-Generated Inventions the Future?

The use of artificial intelligence (AI) and other technologies in the creation of music, art, and literature, as well as in developing novel scientific theories, is not an unfamiliar concept to many of us. From online search functions to mobile phone assistants, AI now pervades many aspects of our lives. For a number of decades, computer-based technologies have assisted in the creation of music and literature.

However, the landscape has been shifting over the last year or two in a way that threatens to challenge some of the core concepts of intellectual property (IP) law. In particular, the increased “autonomy” of AI—whether this is in the creation of artistic works, the development of scientific theories, or the making of decisions about branding and purchasing—raises questions about whether an AI system can itself be an inventor or author of IP rights.

This article explores some of the key areas of interplay between AI and IP law, in particular the possibility of providing IP protection to AI-generated inventions and artistic/creative works. Is this feasible within the existing legal framework, or even desirable? What impact will these developments have on brands across the world?

What Is an AI-Generated Invention?

The last two years have seen the emergence of several high-quality, AI-generated creative works (moving beyond AI-assisted works). David Cope’s Emmy (Experiments in Musical Intelligence) software created realistic melodies in the style of Bach, Beethoven, and Mozart; an Artificial Neural Network has created a sophisticated artwork in the style of Raphael; and an AI program in Japan wrote a short novel that passed the first round of a national literary competition. On the scientific side, the DABUS machine came up with ideas for two potentially patentable inventions: a device for attracting attention for use in search and rescue, and a type of beverage container.

Of course, such “works” are not conceived independently by an AI and are usually the result of “machine learning,” where a person, or people, feed a large amount of data into the AI system (for instance, music composed by Bach, books produced by authors) which it uses to generate novel works in the same style. This mimics the human recombinant creative process—in which we draw inspiration and ideas from existing works to produce new works—but accelerates it exponentially. However, more recently, AI has advanced beyond machine learning to “reinforcement learning,” which has allowed it to become increasingly better at learning and decision-making on its own. For example, DeepMind’s AlphaGo managed to beat the world’s best Go player in 2015 using this type of technology.

IP Rights in AI-Generated Inventions

The question of who should own the IP rights in an AI-generated work is not as simple as it first appears. In Europe, it is a general precept of European IP law that the first owner of an IP right should be the creator or inventor of the work in question. This underlies the entire legal framework which, for example, allows for that person to assign or sub-license the right, or for that right to be transferred automatically to that person’s employer if the act of invention took place in the course of that person’s employment. However, this precept only works in practice because it assumes the creator or inventor is a human with a legal personality, in the sense that the law recognizes the creator or inventor as a party that is and can be subject to rights and duties under the law. A company has legal personality, and so do individual humans, but animals, plants, and machines currently have no legal personality.

What happens when an AI has had sufficient creative input into a work that it can be considered the inventor? Are there any IP rights in that “work” and if so, to whom do they belong?

One solution could be found in existing UK copyright law, which provides that the author and first owner of copyright in a computer-generated literary, dramatic, musical, and artistic work shall be taken to be “the person by whom the arrangements necessary for the creation of the work are undertaken.” Section 9(3) Copyright, Designs and Patents Act 1988. It is conceivable that such a concept could also be extended to other IP rights, such that the rights are granted to the person who makes the necessary arrangements for the AI to, for example, create a design or identify a novel scientific concept.

However, in reality, it may not always be easy to identify that person: it could be the inventor or programmer of the AI, the person who fed existing data into the AI to enable the machine learning process, the person who selects the best or most novel output from the AI, or indeed some combination of all these people. Moreover, this approach still presupposes that one of these people had sufficient input such that they can be considered to have made the arrangements necessary for the creation of the work. But, in a situation in which the AI has effectively acted autonomously, this will not be the case—meaning that, according to this test, there would be no IP rights in the AI-generated “work.” In effect, there would be no one to own the right under this model, as neither the AI machine nor a human would be entitled to the right, leading to the work or invention entering directly into the public domain.

Could the law adapt to allow for AI to be treated instead like an employee-inventor—or even introduce a new provision that the owner of an AI is automatically entitled to any IP rights in its output? Using this model, if the AI created a work, the IP rights in the work would automatically vest in the AI’s owner or programmer rather than the AI itself. This seems like an elegant solution which allows for the AI to be properly recognized as the inventor, but for its owner or programmer to benefit from any IP rights that vest in the AI’s output.

However, this conflicts directly with one of the underlying principles of European IP law: the inventor or creator of a protectable work must be a human. This permeates through European patent, copyright, and trademark law, from the requirement for “skill and labor” or “intellectual creation” in a copyright context, to the definition of “inventor” under the European Patent Convention. It goes back to the foundations of why IP law exists: to protect human creativity. Changing the law on this point would have significant and widespread ramifications for all IP law.

Even if IP law could adapt to recognize a non-human inventor, this solves only half of the problem. An AI does not have a legal personality, it is precluded in any event from being party to any employment agreement, from owning any IP rights, and from being able to transfer any such IP right to a third party (such as its owner).

This was the objection raised recently by both the European Patent Office (EPO) and UK Intellectual Property Office (UKIPO) when the creator of the DABUS machine sought to patent its two inventions. The two patent applications at the UKIPO were refused in December 2019. The creator of the DABUS machine appealed the decision to the English court, but the High Court upheld the UKIPO’s decision in September 2020. The UKIPO commented that “there appears to be no law that allows for the transfer of ownership of the invention from the inventor to the owner in this case, as the inventor cannot itself hold property.” The High Court decision has now been appealed to the Court of Appeal. In January 2020, the two patent applications at the EPO were also refused. The creator of the DABUS machine has also appealed that decision but that appeal has not yet been heard. In the end, to grant AI a legal personality would also be a dramatic departure from existing legal principles and would have vast and serious consequences for all jurisprudence, from contract law to tort.

Perhaps the answer, then, is that truly AI-generated works or inventions should have no IP protection at all. If IP law truly exists to protect and reward human creativity, it is not illogical to conclude that AI-generated works—which have not received at least a de minimis level of human creative input—should be exempt from IP protection. There is even a philosophical argument that protecting AI-generated works would lead to increased outsourcing of human creativity to machines and reduced financial incentives for human creators to learn and develop their crafts.

But this argument runs both ways: refusing IP protection to AI-generated works could also disincentivize people from innovating using AI, in the knowledge that those works could be copied or exploited with impunity. And in a scientific context, as the creator of DABUS has argued before the UKIPO and EPO, exempting AI-generated inventions from patent protection could undermine the aims of the patent system entirely (namely, to encourage scientific innovation).

Conclusions

The remarkable developments taking place in the AI world have vast implications for IP law on a global scale. As AI evolves beyond the need for any significant human input, legislators will need to look very closely at long-established concepts of IP law such as authorship, legal personality, and entitlement, and consider carefully whether and how the law needs to adapt in response to these changes. Key to this process will be remembering and balancing the key principles underlying IP law, such as the need to reward human creativity and the desire to encourage innovation across the artistic and scientific worlds.

On a practical level, brand owners should watch these developments closely because AI that is permitted to own or create works that have IP rights will have a huge effect on how brands operate. For example, if AI-generated works become protectable, this will likely mean an exponential increase in the number of registered rights (whether designs, patents, or trademarks) and a corresponding increase in the need for careful clearance searches before launching a new brand or campaign. It could also open up opportunities for brand owners to invest in AI and reap the rewards from any AI-generated inventions.

Conversely, if legislators choose not to protect AI-generated works, brand owners in creative or technological fields will need to think carefully about how they use any AI in the future and whether such AI needs to be supervised by or work collaboratively with employees to allow any inventive or creative works to attract IP protection. Either way, the impact of legislators’ decisions in this area cannot be overstated.

Although every effort has been made to verify the accuracy of this article, readers are urged to check independently on matters of specific concern or interest. 

© 2021 International Trademark Association