The Dabus case: Artificial intelligence cannot be designated as the inventor in a patent

Thanks to Anna Durante for collaborating on this article.

Artificial intelligence (AI) plays a fundamental role in the life sciences sector. AI offers the opportunity to facilitate various processes. For example, it can be used to analyze and process large quantities of health data that in turn may facilitate the discovery of new drugs, improve diagnostics, and customize treatments. It also can be used to automate laboratory processes and manage clinical trials, saving time and resources.

However, such use also raises numerous ethical and legal questions that touch on responsibility, reliability, and the correct and legal use of data and information to train AI for new technical solutions. These issues have been the subject of wide-ranging discussion and high-profile legislative maneuvers, including the recent AI Act.^[1]

An open issue

Many issues remain unresolved. Use of AI also raises questions about the patentability of discoveries generated by AI and the possibility of naming AI as the author of inventions subject to patent protection.

Can an invention generated by AI be the subject of a patent? If a new technical discovery is generated by AI, who should be named as the inventor? What role does human intervention in the form of guiding, selecting, and establishing data play in generating the invention and how does that work for the purposes of attributing authorship of the invention?

These and other questions are involved in the saga known as the Dabus case, which has traveled through patent offices and courts in various countries around the world for years. Recent developments in the case may influence the future of patentability of AI-generated inventions.

The Dabus case in brief

The Dabus case centers around Stephen Thaler, a scientist and the CEO of Imagination Engines Inc., a neural networks company based in Missouri. Thaler is the creator and owner of Dabus, an acronym for device for autonomous bootstrapping of unified sentience. Dabus is an AI system designed to generate creative outputs autonomously. It does so with little direct human intervention, limited to administering training data.

For years, Thaler has argued that Dabus is capable of autonomously conceiving original inventions. He is determined to claim Dabus as an inventor in patent applications for discoveries generated through the system and filed in various jurisdictions.

Thaler argues that the inventive process that led to the technical solutions that are the subject of the patent applications was carried out independently by AI and, consequently, the invention must be patented with Dabus designated the sole inventor instead of assigning ownership of the patent to Thaler as the owner of the AI system.

AI cannot be designated as inventor

Despite Thaler’s assertion of Dabus’s “inventive capacity,” most national patent offices, including those of the United Kingdom, Australia, and the United States (and with the exception of that of South Africa, though the South African patent office performs no substantive examination of an application and simply determines compliance with certain formal requirements), have rejected Thaler’s patent applications. They determined that under current laws an inventor must be a natural person and not a machine. Since a machine does not have legal capacity, it cannot claim any individual rights and therefore cannot be designated the inventor in a patent.

The European Patent Office (EPO) also stated that under the European Patent Convention (EPC) only a human being can be considered an inventor and designated as such in a patent application as a party with legal capacity.^[2]

Ownership of the patent has also been excluded for the owner of Dabus

The EPO also rejected Thaler’s auxiliary request to designate himself, as the owner and creator of Dabus, as the patent holder. According to the EPO, current regulations state two principles:

• Upon filing, the European patent application must indicate the natural person designated as the inventor, a requirement that was not met in the Dabus case;
• Even if this rule did not apply, when the applicant (Thaler) and the inventor (Dabus) differ, a declaration as to the origin of the applicant’s right to ownership of the European patent is required. The declaration submitted by Thaler that he is entitled to the European patent as the owner and creator of Dabus is not sufficient because it does not cite a legal situation or legal transaction that would have made him a successor in title to the inventor, since the AI cannot be the owner of any right and therefore cannot assign any right under the EPC.

The UK position

The same position was taken by the Supreme Court of the United Kingdom in December 2023. At the end of a long and hotly contested proceeding initiated by Thaler before the British Intellectual Property Office, it ruled on the same issue.

According to British judges, the patent applications filed by Thaler naming the AI system as inventor do not meet the requirements of UK law, which states that the inventor can only be a natural person and therefore cannot be an AI system. The Supreme Court of the United Kingdom concluded that:

• The only reasonable interpretation of the applicable law is that the inventor must be a natural person, as the possibility that a machine acting alone and operated by AI could be an inventor is not contemplated;
• Regarding Thaler’s right as owner of the machine to apply for a patent on technical advances made by Dabus, UK law provides that a patent may be granted to the inventor or to a person acting in place of the inventor;
• In this case, similar to the EPO, the UK court found that Dabus was not an inventor, nor could it assign rights to the invention to Thaler, since it does not have legal personality;
• Finally, regarding the possibility of invoking the principle that the owner of an object is also the owner of its fruits, according to the Supreme Court, no such rule exists in relation to intangible rights such as intellectual property.

Therefore, under the current law there was no room for patentability for inventions generated by AI.

While there is a growing need to find a form of protection for inventions generated by AI, the united front of decisions has begun to show weakness in recent months.

A potential opening: The role of the “significant inventive contribution” of the AI system user according to the U.S. Patent Office

In February of this year, the U.S. Patent Office (USPTO) published guidelines on the patentability of inventions generated by AI. This marked a small but significant step forward from previous decisions.

The USPTO confirmed that an AI system cannot be designated as an inventor, but it also recognized that inventions resulting from the use of AI systems are not excluded ex ante from patentability. Indeed, such inventions are certainly patentable when the natural person who used the AI system has made a “significant inventive contribution” to identifying the new technical solution. In that case, the natural person will be referred to as the inventor.

USPTO guidelines outline the criteria necessary to determine whether requirements that integrate the concept of a “significant inventive contribution” have been met.

Data isn’t enough: The human being must play a significant role in arriving at the inventive solution

The human being must play an active and decisive role in devising the invention and cannot merely provide data or present a problem to the software. Identifying a problem or positing a general goal or research plan to pursue is not sufficient for conception and thus inventive contribution.

A natural person who merely presents a problem to an AI system cannot be an inventor or co-inventor of an invention resulting from the AI system’s output.

However, the original and innovative way a person constructs a prompt about a specific problem to elicit a solution from the AI system could be a “significant inventive contribution.” An individual who merely recognizes and appreciates the output of an AI system as an invention, particularly when the properties and utility of the output are obvious to those with standard knowledge of the field, is not necessarily an inventor.

Still, a person who takes the output of an AI system and makes a significant contribution to it to create an invention, e.g., by selecting and coordinating the results obtained, may be referred to in a patent as a proper inventor.

Inventive contribution: The ruling of the German Federal Court of Justice

On June 11, 2024, a further step was taken in the Dabus saga: the German Federal Court of Justice concluded a long series of proceedings initiated by Thaler before the national patent office with a ruling that offers interesting thoughts on AI-generated inventions.

The court confirmed the firmly established stance in almost all jurisdictions that only a natural person, and not an AI system, can be designated as the inventor in a patent application. But it also offered further reflection on the human contribution required to recognize an AI-generated product as a patentable invention.

During the first-instance proceedings before the Federal Patent Court, Thaler submitted ancillary claims to the main claim to designate Dabus as an inventor in patent applications. He proposed himself as the inventor, as he was the natural person who had created the prompt that allowed Dabus to generate the invention.

In 2021, the Federal Patent Court decided to grant a patent on the basis of this auxiliary claim, designating Thaler as the inventor with the addition of the phrase “who induced the artificial intelligence Dabus to generate the invention.” The decision was appealed by the German Patent Office, but the recent decision of the Federal Court of Justice upheld the ruling, thus opening up more possibilities to register inventions based on the use of AI.

Significant human intervention is enough

In keeping with the USPTO guidelines, the court ruled that it is possible to designate a natural person as the inventor of a discovery generated using an AI system. However, the court went beyond those guidelines, stating that the role of inventor does not necessarily require an independently inventive contribution, but can derive from any significant human contribution that has in some way influenced the overall success of the invention.

Based on current scientific knowledge, the German court asserted that no AI system could draw on technical teachings without human preparation or influence, and such human preparation or influence is sufficient to determine that the natural person behind it fills the role of inventor.

In essence, according to the Federal Court of Justice, behind generation of an invention—even an AI-generated invention—there is always a human being who had significant influence. The type of influence is not important, and the human contribution does not have to be inventive or creative.

In other words, if an appropriate problem is posed to an AI system and the AI system solves the problem by creating an invention, the person who provided the instructions and posed the problem to the AI system can be named the inventor in the relevant patent.

Future perspectives

The recent decision of the German Federal Court of Justice in the Dabus case marks an important step in recognizing the role of AI in technological development. It outlines a possible path for the protection of inventions generated in this way.

What impact will this decision have on future cases in which national patent offices have to decide on the patentability of an AI-generated invention? Will the German court’s approach be adopted in other countries? This seems highly likely in Europe, where patent law is already largely uniform—even more true with the establishment of the European patent and the Unified Patent Court becoming operative last year.

What is certain is that there is an urgent need to find a solution for protecting inventions generated through the use of AI. This need is pressing, and a solution must be reached, either through case law or, ideally, with clear legislative action on the matter.

[1] European Regulation on Artificial Intelligence No. 1689/2024.

[2] In its 2022 decisions on cases J 8/20 and J 9/20.