In recent years, AI technologies have gained enormous importance and fundamentally changed our everyday lives through groundbreaking applications. But what about the protection of intellectual property when it comes to AI inventions? Are AI patents even possible under current patent law? In this blog post, we take a closer look at the key questions of whether AI can be an inventor or author, whether it has legal capacity, and how AI inventions can be protected by patents.
Do you need assistance with filing an AI patent? Our experienced software patent attorneys will be happy to support you. Contact us for an initial consultation!
Table of contents
- https://rgth.de/2024/09/ki-patente/#was-ist-kiBackground: What is AI?
- https://rgth.de/2024/09/ki-patente/#patente-fuer-kuenstliche-intelligenzPatents for Artificial Intelligence
- https://rgth.de/2024/09/ki-patente/#erfinderische-taetigkeit-bei-ki-erfindungenInventive step in AI inventions
- https://rgth.de/2024/09/ki-patente/#der-comvik-ansatzThe COMVIK approach in case law
- https://rgth.de/2024/09/ki-patente/#kann-eine-ki-erfinder-seinCan an AI be an inventor?
- https://rgth.de/2024/09/ki-patente/#rgth-patentanwaelte-fuer-kuenstliche-intelligenzRGTH: Patent attorneys for Artificial Intelligence
Background: What is AI?
Artificial intelligence has been part of many areas of our lives for years. Simple examples of AI can be found online: search engines, spam filters, translation tools, text‑generation tools, or recommendation systems—whether for purchasing decisions or the next series on Netflix & Co. Further examples include autonomous driving, virtual reality (VR) and augmented reality (AR) creating digital worlds, navigation systems that guide us along the most efficient routes, or smart‑home solutions that control electricity, heating, or water consumption. Most recently, AI applications based on large language models such as ChatGPT have attracted widespread attention and become accessible to the general public.
All of these AI solutions have one thing in common: they are based on artificial neural networks and algorithms that are trained using data to recognize, analyze, and interpret patterns. On this basis, AI applications are capable of making independent decisions and performing intellectual tasks such as logical reasoning, learning, or problem solving.
The innovation with ChatGPT is that users can now “converse” with AI in natural language via a chat interface—without requiring in‑depth coding knowledge.
The difference between “strong” and “weak” AI
When looking at the AI applications we know today, we generally speak of weak AI. Computers and machines automate tasks, improve efficiency and accuracy in various fields, and can partially emulate human cognitive processes. They are trained to perform specific tasks and can therefore only operate effectively within the domains for which they were developed.
What weak AI lacks is creativity, general understanding, and consciousness. AI systems possessing these characteristics would be able to solve intellectual tasks at a level comparable to humans, adapt to new situations, and make independent decisions. Such systems would be referred to as strong AI. While strong AI is already a reality in science‑fiction films, it does not yet exist in practice. It remains a theoretical concept and a long‑term goal of AI research.
Patents for Artificial Intelligence
Weak AI is based on computer‑implemented inventions, often implemented through software. Such solutions can only be protected by patents under certain conditions. According to the https://www.dpma.de/patente/patentschutz/schutzvoraussetzungen/schutz_computerprogramme/index.htmlrequirements of the German Patent and Trade Mark Office (DPMA), patent applications for mathematical methods and computer programs are only permitted to a limited extent. The European Patent Convention (EPC) excludes mathematical methods, plans, rules, and methods for mental acts from patent protection when claimed “as such”.
The relevance of this topic is underscored by the growing number of AI patent filings worldwide. China leads the field with more than 38,000 patents in the area of generative AI over the past ten years. The United States recorded 6,300 patent filings, and Germany 708. Particularly noteworthy is that more than one quarter of these applications were filed in the past year alone.
Our experienced patent attorneys for Artificial Intelligence will be happy to assist you with the https://rgth.de/rechtsgebiete/patente-anmelden/filing of patents.
When are AI inventions patentable?
To obtain patent protection for AI solutions, the invention must use technical means to contribute to the solution of a specific technical problem. This means that computer programs based on AI inventions are patentable if the invention:
- goes beyond a purely abstract or intellectual concept
- is carried out using a computer, computer network, or another programmable device with a technical effect, and
- has a technical character.
In addition, patents can only be filed for final AI applications. Individual development stages of AI‑based solutions are not patentable. Furthermore, AI inventions can only be the subject of a patent if they meet the general patentability requirements of the European Patent Office (EPO): they must be novel, involve an inventive step, and be industrially applicable.
Inventive step in AI inventions
Although the patentability of AI is recognized in the EPO examination guidelines, assessing whether AI patent applications will be successful is often challenging. In principle, an inventive step exists if the invention does not arise in an obvious manner from the prior art. However, the inventive step is frequently called into question by the state of the art.
The assessment is based on the knowledge of a skilled person—one or more experts with average knowledge and abilities in the relevant technical field who are aware of what constitutes common general knowledge at the relevant time.
It is sometimes assumed that the use of AI leads to a broader body of prior art. The more powerful AI becomes, the more it will be used as a standard tool, and the more patent applications are likely to be filed in this field in the future. This could further complicate positive assessments of novelty and inventive step. At the same time, the question arises as to whether AI will soon be creative enough to generate subject matter that meets the high requirements of inventive activity.
These considerations are relevant not only for patent applications, but also for the https://rgth.de/rechtsgebiete/gebrauchsmuster-anmelden/filing of utility models or when applicants wish to protect a https://rgth.de/rechtsgebiete/geschmacksmuster-anmelden/design.
The COMVIK approach in case law
To simplify assessment, case law increasingly applies the so‑called “COMVIK approach”. Under this approach, inventive step is assessed by considering only those differences from the closest prior art that contribute to the technical character of an invention. Both technical and non‑technical features are taken into account, provided that they individually or in combination solve a technical problem.
Can an AI be an inventor?
In addition to the question of whether AI is patentable, ongoing technological progress raises the issue of whether AI itself can be regarded as an inventor.
In principle, three categories of AI‑related inventions can be distinguished:
- Inventions developed by AI that are independently capable of identifying and solving tasks.
- Inventions based on human intelligence and reviewed with the assistance of artificial intelligence.
- Inventions in which humans use AI to solve a specific task.
The scientific consensus is that inventions of the first category—those created independently of humans—remain a vision of the future. Accordingly, under current law, only a natural person may be named as inventor in patent applications for AI inventions. AI itself cannot apply for IP rights and is not regarded as an independent legal entity, but merely as a tool used by the inventor.
AI as an inventor: current case law of the German Federal Court of Justice
In June 2024, the German Federal Court of Justice (https://juris.bundesgerichtshof.de/cgi-bin/rechtsprechung/document.py?Gericht=bgh&Art=en&nr=138469&pos=0&anz=1BGH, X ZB 5/22) ruled that only natural persons may be recognized as inventors. Naming an AI as the sole inventor violates Section 37(1) of the German Patent Act. A human contribution therefore remains mandatory—even if artificial intelligence assisted in developing the invention. References to the use of AI may be included as supplementary information, but the designation of a human inventor is compulsory.
Outlook: AI and its legal recognition
Under current case law, AI does not have legal capacity. However, this does not mean that machine‑generated inventions developed with the assistance of AI are excluded from patent protection. At present, however, IP rights may only be claimed by natural persons.
The legal situation may change if AI research succeeds in developing weak AI into strong AI and transforming the theoretical concept into reality. Should AI develop creative and inventive capabilities, this would raise the question of whether it possesses its own consciousness and should be legally equated with human intelligence. In that case, the assessment of whether AI has legal capacity—and thus can be an inventor, author, or patent applicant—would need to be reconsidered.
RGTH: Patent attorneys for Artificial Intelligence
Do you need assistance with filing an AI patent? Our software patent attorneys specialize in artificial intelligence, computer‑implemented inventions, and cloud computing, and support you in all matters relating to patent and IP protection for AI inventions. From assessing patentability and developing a patent strategy to filing the patent application, our patent attorneys accompany you throughout the enforcement of your IP rights.
Disclaimer: The information provided above does not constitute legal advice. No guarantee is given as to its accuracy or completeness.