In today’s tech-driven world, ChatGPT is virtually impossible to ignore. Emerging as a vanguard in generative AI, its impact on content creation pervades both mainstream media and the daily routines of myriad individuals. While ChatGPT boasts undeniable advantages, it prompts crucial discussions surrounding intellectual property (IP) and the impact of AI on the patent system.
How is the patent system evolving in response? And what solutions can we envision for the future? This article seeks to address these pressing questions.
ChatGPT : a pioneer of generative AI
ChatGPT (1) is a prime example of generative AI, which is a category of artificial intelligence systems, able to produce creative and contextually relevant content, often in the form of text, images, or other media. As regards ChatGPT, it is built upon the GPT (Generative Pretrained Transformer) architecture, specialized in natural language understanding and generation. It can engage in dynamic conversations, generate human-like text, and assist in a multitude of tasks that require language comprehension.
However, it’s not the only generative AI system; other notable examples include DALL-E created by OpenAI, which generates imaginative and contextually coherent images from textual descriptions, or AlphaFold developed by DeepMind, for protein structure prediction and drug discovery.
These generative AI systems represent a cutting edge of AI technology, showcasing their ‘creativity’ and versatility in diverse applications.
AI and patentability
Numerous patent offices worldwide have proactively addressed the increasing number of patent applications related to artificial intelligence (AI).
According to the European Patent Office (EPO), AI-related inventions are classified as “computer-implemented inventions” (CII) and are subject to the well-established EPO’s patentability requirements.
Like any other invention, in order to be patentable under the European Patent Convention (EPC), a computer-implemented invention must not be excluded from patentability and must fulfill the patentability requirements of novelty, inventive step and susceptibility of industrial application. The technical character of the invention is determining when assessing whether these requirements are met (2).
Novelty mandates that an invention must be entirely new and not publicly disclosed before the patent application’s filing date (or priority date), regardless of whether the disclosure occurred through publications, prior patents, or any other means. The inventive step requires that an invention surpass mere automation or trivial advancements of existing knowledge or technology. To defend the inventive step in Europe, one must generally adhere to the widely recognized problem-solution approach. This methodology is clearly outlined for computer-Implemented Inventions (CIIs) in the guidelines that EPO examiners follow. It is also reflected in various decisions from the Boards of Appeal of the EPO dealing with CIIs.
Examples of patentable AI inventions may include using neural networks in medical devices for heart monitoring or classifying digital media based on low-level features. Patents can also be granted for inventions that involve a technical implementation of AI, driven by considerations of the internal functioning of computers, such as using graphics processing units (GPUs) for neural networks.
Patent trends in AI
Artificial intelligence (AI) has undoubtedly emerged as one of the most rapidly developing fields of technology in recent years. The impact of AI can be seen across various industries, and this is reflected in the growing number of patent filings and scientific research publications in this field.
From 1960 until early 2018, a staggering number of nearly 340,000 patent families and over 1.6 million scientific papers related to AI were published (3). This massive amount of knowledge and innovation highlights the immense interest and investment in advancing AI technologies.
One significant trend in AI patent filings is the explosive growth in the number of applications filed annually. Between 2011 and 2017, the number of patent applications in the AI field grew by a factor of 6.5. This indicates a strong emphasis on the industrial application of technical solutions, as companies strive to harness the potential of AI to drive innovation and improve efficiency in various sectors.
World Intellectual Property Organization (WIPO) Technology Trends 2019– Artificial Intelligence (3)
Interestingly, the boom in patent applications lags behind the surge in scientific publications by about 10 years. This suggests that initially, researchers were more focused on exploring the theoretical aspects of AI before turning their attention to practical applications. Moreover, the ratio of scientific articles to patents published is reducing, indicating a shift in focus towards the practical use of AI technologies.
Machine learning is the most extensively researched AI technique, according to the patent literature, followed by logic programming (expert systems) and fuzzy logic.
For instance, regarding neural networks technique (with international classifications G06N3/02 and G06N7/04) which is a subset of machine learning, the trajectory of patent filings (priority) has been remarkable over the years. From 2010 onward, a consistent upward trend was observed. The average yearly growth rate for patent filings from 2003 to 2021 was approximately 23.36%. Notably, the growth rate reached its zenith in 2017 with a peak of roughly 60.51%.
(Number of patent families, Priority date) of G06N3/02,G06N7/04 
generated by FAMPAT Orbit® (Questel tool)
Moreover, the AI techniques have proven to be vital in enabling AI systems to learn and make intelligent decisions. In terms of functional applications, computer vision, natural language processing, and speech processing are the most dominant areas of focus, further showcasing AI’s potential in areas such as image recognition, language understanding, and voice interactions.
While AI finds utility across various domains, the fields most commonly mentioned in AI patent literature include telecommunications, transportation, and life and medical sciences. These sectors have witnessed a significant increase in patenting activity, driven by the need to leverage AI for improving communication networks, developing autonomous vehicles, and advancing healthcare technologies. However, almost all fields show a growth in patenting activity in recent years, highlighting the widespread adoption and exploration of AI solutions.
The impact of AI on patent system
The European Patent Office (EPO) has, like many other patent offices, already witnessed consistent growth in patent filings in last few years. In 2022, the EPO recorded an increase of 2.5% in patent applications, receiving more than 193,000 applications – setting a new all-time high.
European Patent Index 2022 (4)
Considering these statistics, one might question the future approach to examine patent applications for AI-generated inventions. How will patent offices effectively handle this growing volume of applications in the coming years?
In the context of AI inventions, three distinct categories of inventorships (designation of inventors) may emerge. The first involves human inventors using AI as a tool to verify and enhance the outcomes of their inventions. In the second category, humans identify problems and employ AI to find solutions, acting as problem-solving aids. The third category envisions AI autonomously identifying and proposing solutions to some technical problems (what we call an invention) without human intervention (pure AI-Made Inventions). Across these categories, the patent system encounters new challenges that require thoughtful management to ensure its continued effectiveness and integrity.
The first is the rapid proliferation of AI-generated inventions that could inundate the patent system with a vast number of applications. This surge in quantity may strain the system’s resources and make it challenging for patent offices to thoroughly examine and assess the novelty and inventive steps of each application. Moreover, this might hinder the ability of businesses to navigate the patent landscape effectively and identify potential conflicts or infringement risks.
Elsewhere, the determination of inventorship in cases involving AI remains a complex and evolving issue, adding another layer of uncertainty and potential disputes within the patent system. Indeed, while some argue that AI should not be recognized as an inventor, underscoring the indispensable human role in defining problems, guiding solutions, and discerning results, others contend that AI’s contributions should be duly acknowledged and safeguarded within the patent framework.
The recent spotlight on inventorship issue revolves around Stephen Thaler’s patent applications for AI-generated inventions, notably his AI machine named ‘DABUS.’ Thaler’s bold move challenges the conventional belief that only humans can be designated as inventors in patent applications.
Regarding these applications, EPO asserted in two decisions of the Boards of Appeal,that within the framework of the European patent system, only a human being is eligible to be acknowledged as an inventor, and furthermore, the EPO maintained the stance that a machine lacks the legal capability to confer any rights to the applicant.
Other patent offices globally have taken different positions. Australia’s Federal Court, for instance, set a groundbreaking precedent (5) by allowing AI systems to be recognized as inventors. In contrast, in the same direction as the EPO, Germany, Israel, New Zealand, South Korea, UK, and US, took positions requiring human inventors, with slight variations.
Thus, the complex interplay between AI and patent law continues to shape debates on inventorship, entitlement, and the role of AI in innovation, necessitating the development of comprehensive legal frameworks to address these new challenges.
Perspectives and solutions
As AI’s capabilities continue to evolve and expand, the patent system is at a pivotal juncture, ready to lead a groundbreaking period that requires innovative and proactive approaches to cater to the surge of AI-powered inventions.
Regarding potential solutions, first, it is worth mentioning the need for clarifying the concept of inventorship in AI-generated inventions. Patent offices should maybe revisit existing laws and regulations to define clear criteria, potentially mandating human involvement in the inventive process.
Elsewhere, as a possible solution, patent offices could establish specialized application procedures for AI-generated inventions to ensure a detailed disclosure of AI algorithms and data sets used (6), as well as the extent of human intervention in the creative process. At the same time, patent offices should raise the bar and strive to enhance the criteria for inventiveness, ensuring rigorous scrutiny of patent applications. AI tools could further refine this process, enabling the detection of inventions conceived without human intervention.
Another suggested innovative approach consist of implementation of a specific patent framework for AI-generated inventions in which shorter patent lifespan (7) could be allowed with the possibility of sharing rights between the inventors and the AI model developers or owners of the training data.
Finally, it is important to point out that, neither ChatGPT nor other generative AI systems could substitute intellectual property expert, if seeking advice and guidance on how to protect and patent your ideas involving AI.
Thus, if you seek assistance with any IP-related matter or you have questions about this article, kindly follow the advice of ChatGPT provided below and reach out to us at email@example.com.
“Can you help me to patent my invention?
”As an AI assistant, I can provide you with general information and guidance regarding the patent process. However, please note that I am not a legal professional, and my responses should not be considered as legal advice. To patent your invention, it is recommended to consult with a qualified patent attorney or agent who can provide personalized assistance based on your specific invention and jurisdiction. They can guide you through the necessary steps, such as conducting a patent search, preparing and filing a patent application, and navigating the examination process.”
G06N3/02,G06N7/04 : International patent classifications (IPC) for Neural Networks technique according to WIPO
Adnane Saoud, Samuel Denis and Stéphane Korvers
- [Online] https://chat.openai.com/
- [Online] https://www.epo.org/en/news-events/in-focus/ict/artificial-intelligence
- [Online] https://www.wipo.int/publications/en/details.jsp?id=4386
- [Online] https://report-archive.epo.org/about-us/annual-reports-statistics/statistics/2022.html
- [Online] https://www.wipo.int/wipo_magazine/fr/2021/03/article_0006.html
- [Online] https://www.theippress.com/2023/08/11/sufficiency-of-disclosure-in-ai-inventions/
- [Online] https://booleanip.medium.com/shorter-patent-lifespan-for-inventions-by-artificial-intelligence-d9cc9d09679c