I. Introduction

In recent years, Artificial Intelligence (AI) has evolved very rapidly in terms of performance and has revolutionized numerous fields, from medicine to finance, from industrial production to artistic creativity. In this context of very rapid technological evolution, in parallel with the rapid growth of AI applications, the debate about patents related to these technologies has developed, raising important questions about intellectual property, protection of innovation, and ethical and legal challenges.

Intellectual property, and patents in particular, are one of the key tools for protecting technological innovations. The patent system grants patent holders exclusive rights with territorial value for a limited period of time. The protection provided by patents allows patent holders to obtain a return on investments made in research and development (R&D) to incentivize further investments- It also prevents third parties from exploiting their innovations without compensation. However, AI - with its innovative features and inherent potential - is challenging the international patent system.

Patenting solutions based on AI technology is complex, as many of the core elements of AI involve machine learning algorithms, mathematical models, computing processes and neural networks, which in many jurisdictions are not patentable per se. To circumvent this obstacle, methods of using the algorithms in specific applications or AI-related hardware solutions are generally patented.

In recent years, there has been a boom in the number of AI-related patent applications filed, which grew exponentially between 2015 and 2022. China and the United States lead the list of largest filers, followed by Europe, Japan, and South Korea.

II. Main sectors in which AI is involved

The main areas where patented innovations concerning AI are focused on include:

Machine learning: patented technologies include methods for improving the efficiency of algorithms, such as neural networks, and specific applications like speech or facial recognition;

Medical sector: in this field, many filed applications concern the use of AI to improve diagnosis and treatment, such as for analyzing medical images, predicting the effectiveness of treatments, or discovering new drug combinations;

Autonomous vehicles: AI is present in many patent applications concerning autonomous driving, particularly applied to perception, control and navigation systems;

Robotics and automation: many innovations in this area involve the use of AI to improve the ability of machines to interact with their surroundings and make decisions autonomously;

Finance and investment: several patents (in jurisdictions that allow them) cover innovations ranging from improving the security of payment platforms to using AI to manage financial investments.

III. Patentability of AI-related technologies

The patentability of AI-related technologies is a controversial topic, as the intangible and often abstract nature of the elements underlying innovations based on or exploiting AI makes it difficult to establish clear boundaries.

The European Patent Convention (EPC) allows the EPO to grant patents for inventions in many fields of technology where AI finds technical application. Specifically, the European Patent Office (EPO) follows the same approach for evaluating AI-related innovations as it does for computer-implemented innovations (CII) (EPO Guidelines, G-II, 3.3.1).

As a result, although AI is based on computational models and mathematical algorithms that are themselves abstract in nature, patents can be granted when AI moves out of the abstract scope and is applied to solve a technical problem in a technological field. Examples of innovations that are considered patentable can be found in EPO Guidelines G-II, 3.3). For instance, a machine learning algorithm used to improve the performance of a medical device might be patentable, while the algorithm itself is not. In general, one point to consider would be crafting claims to emphasize technical effect or specific applications rather than the underlying algorithms. Indeed, before the EPO it is the technical aspect of the part of the invention that is deemed to be new that concurs to the assessment of inventive step.

Even in the United States, the U.S. Patent and Trademark Office (USPTO) and the Courts have repeatedly emphasized that simple mathematical formulas or abstract patterns cannot be patented unless they are innovatively applied to a concrete technical problem. 

Further, it shall be mentioned that AI tools are used collaboratively across different entities. In such cases, data ownership issues may arise in AI-generated inventions, as the training data for AI systems is often sourced from multiple entities. This would be a significant consideration in determining who owns the output of an AI system. 

In this context, introducing contracts and licensing strategies for companies developing AI tools is now part of best practice, ensuring clear ownership and use rights for AI-generated inventions.

Further, potential risks of patent trolls leveraging AI-generated inventions could lead to excessive litigation and hinder smaller innovators. This ties into the broader discussion of AI potentially becoming a double-edged sword in the innovation landscape.

IV. Rights of inventors and inventions created by AI: DABUS case

Another complex and controversial issue concerns the protection of the rights of creators of AIs and inventions obtained using AI systems. In recent years, there has been particular debate as to whether AI systems can be considered as inventors of a patent. The debate has become particularly heated with the DABUS case, which has led to a series of rulings and decisions by various Patent Offices on this issue.

DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) is an AI system created by Dr. Stephen Thaler. DABUS allegedly independently generated two inventions, one concerning a food container based on fractal geometry, and another concerning a flashing beacon to attract attention in an emergency, and its creator filed respective patent applications on the two innovations in a plurality of Patent Offices, for which DABUS was listed as the sole inventor.

The reactions of the various Patent Offices and Courts called upon to decide whether or not an AI system could be recognized as an inventor for the purposes of patent law have varied although most commonly aligned in denying that possibility.

In Europe, the EPO rejected patent applications in which DABUS was named as the inventor stating that, according to the European Patent Convention, the inventor must be a human being. The Legal Board of Appeal upheld this decision, pointing out that only a natural person can be recognized as an inventor.

In the United States, the USPTO rejected applications filed by Stephan Thaler arguing that an invention must be attributed to a natural person. The Supreme Court upheld this position, holding that DABUS being a machine cannot have legal rights and cannot be recognized as an inventor.

In the United Kingdom, applications were also rejected, and the U.K. Supreme Court ruled that under the 1997 Patent Act, the inventor must be a natural person.

In Australia, the designation of DABUS as an inventor was initially accepted, but this decision was later overturned. The Federal Court ruled that an inventor must be a natural person, aligning with the positions of other patent offices.

South Africa was the only country to grant a patent with DABUS as inventor.

In Germany, the ruling issued by the Federal Patent Court confirmed that at the German Patent and Trademark Office (DPMA), an AI system cannot be considered an inventor under patent law, although applicant Stephen Thaler, who insisted that he did not invent the food container, was suggested to be named as an inventor in the patent application.

V. Issues still open after the DABUS case

The DABUS case and the decisions stemming from it have sparked significant debate about the need to update patent laws to account for new technologies and address the challenges posed by AI.

While most Patent Offices require the inventor to be a natural person, there is no control over the correctness of inventors' names provided other than in formal terms. As a result, inventions developed by AI can be attributed to any human being.

This, on the one hand, would contradict the principle of protecting human creativity that underlies an inventor's moral right to be named as such. On the other hand, it would allow individuals to attribute merits to themselves that are not real and earn credits that are not derived from their own work. Consequently, the work of someone who simply asks an AI to solve a certain problem, without providing any other input, would be placed on the same footing as the work of someone who has actually independently invented something new.

Based on these considerations, designating an AI as an inventor would allow the moral rights of real human inventors to be protected.

Thus, the discussion on the issue remains open. As the use and integration of AI increases in new areas, it can be assumed that the number of patents and applications related to AI is likely to grow rapidly in the coming years. This could result in new regulatory challenges for the various Patent Offices and consequently the need to update current regulations to accommodate new technological developments so as to ensure both the protection of inventors and the promotion of innovation.

It shall also be considered that the risk exists of patent thickets or over-patenting, especially in fast-evolving fields like AI. An increase in AI patent filings could lead to potential litigation issues or monopolistic control, which might stifle smaller innovators.

Also addressing joint ownership, data rights, and licensing issues specific to AI systems appear to become more and more relevant issues to be tackled by companies and professionals.

The debate over AI and patent laws is still unfolding globally, and it would be desirable that more attention is given to potential harmonization of global patent laws to address AI issues. In this regard, efforts by WIPO (World Intellectual Property Organization) and/or regional harmonization efforts would be useful.

To know more about GLP, please visit the website: https://www.glp.eu/en/