I. INTRODUCTION
Patent law is a significant field that ensures the protection of technical inventions and promotes innovation. While the use of artificial intelligence in the legal domain is rapidly increasing, its role in highly technical and specialized areas such as patent law has yet to be clearly defined. Large language models, which learn from extensive datasets to provide language-based analyses, still face challenges in substituting for genuine technical expertise. EPO has delivered an important decision in this respect, discussing that AI-based tools such as ChatGPT cannot be directly accepted as evidence in the interpretation of patent claims. In patent law, the accurate interpretation of technical terms is of critical importance in assessing novelty and inventive step. Traditionally, such interpretation is grounded in the perspective of the “person skilled in the art.” This concept was developed to ensure the objectivity of decisions and is indispensable for the accurate understanding of technical content. In decision T 1193/23, the EPO reaffirmed the importance of the person skilled in the art and emphasized that artificial intelligence should be employed solely as a supportive and auxiliary tool in this context.
II. THE CONCEPT OF THE “PERSON SKILLED IN THE ART” IN THE INTERPRETATION OF PATENT CLAIMS
In patent law, one of the central concepts in assessing whether an invention is patentable is a hypothetical construct known as the “person skilled in the art,”1 also referred to as the “relevant skilled person” in the field. This figure is not a real individual; rather, it is a reference point embodying standard knowledge and skills in the relevant technical field, endowed with rational thinking, the ability to combine existing technical knowledge, and the capability to solve problems within the framework of general knowledge.
The concept of the person skilled in the art in patent law serves as a fundamental criterion for objectively interpreting patent claims and evaluating the novelty and inventive step of an invention. This hypothetical individual is defined as an expert who, at the filing date of the patent, possesses the average level of knowledge and technical capability in the relevant technical field2.
With these attributes, the person skilled in the art represents the average technical level in the interpretation of patent claims. Thus, the scope of patent protection is determined objectively within the general understanding of the technical community. The EPO regards the person skilled in the art as an average expert in the field at the relevant time, one who is well-versed in the state of the art but lacks inventive capacity3. This individual is assumed to be in a state of continuous professional development and therefore is not confined solely to existing knowledge, but remains receptive to sectoral evolution and technical advancements.
The person skilled in the art constitutes a fundamental benchmark for claim interpretation in patent law. The meaning and scope of claims are assessed objectively from the perspective of this figure. This approach both clarifies the scope of protection for the patent proprietor and provides third parties with predictability regarding the limits of protection4. By applying this standard, concepts are concretized and an agreed-upon objective assessment is achieved5 Indeed, to enable such objectivity, the person skilled in the art, assumed to possess average skill in the field, is taken as the reference point.
While there is broad international consensus on the definition of inventive step, there is no uniform agreement regarding the specific attributes the person skilled in the art must possess6. Academic doctrine has advanced various views, suggesting that such a person should be: aware of domestic and foreign publications, informed of developments, knowledgeable in adjacent fields in addition to their own, proficient in the relevant technical area, and capable of keeping up with innovations, among other characteristics7. In summary, the essential features of the person skilled in the art may be listed as: (i) a hypothetical persona; (ii) a qualified practitioner in the relevant field; (iii) possessing average knowledge and skills; (iv) aware of the common general knowledge in a specific technical field as of the relevant date; (v) having access to all information in the state of the art; (vi) possessing routine working capacity but lacking inventive ability; (vii) continuously following developments; and (viii) in some cases, being represented not by an individual but by a team8.
Since technical knowledge evolves constantly, the knowledge framework of the person skilled in the art also changes over time. In interpreting patent claims, the technical level at the filing date of the application is taken into account. This historical perspective ensures that the assessment is based on the knowledge and practices available at that time, thereby preventing the claims from being evaluated in light of subsequent technical developments and ensuring legal certainty.
The legal significance of the person skilled in the art becomes particularly prominent in patent examination and opposition proceedings. Courts and patent offices interpret the scope of claims from the perspective of the person skilled in the art, thereby ensuring that interpretations are grounded in objective technical criteria rather than subjective perceptions. In light of recent technological advancements, the concept of the person skilled in the art has also begun to be discussed concerning AI-based tools; in this regard, the question has arisen as to the extent to which such systems can be taken into account in technical assessment processes.
III. BACKGROUND AND SUBJECT MATTER OF DECISION T 1193/23
Decision T 1193/23, rendered by the EPO Board of Appeal, constitutes an important precedent concerning the patentability of methods for the safe starting and/or stopping of rotor spinning machines. The decision is particularly significant with respect to the criteria applied in assessing the technical content of claims, the application of the concept of the “person skilled in the art,” and the admissibility of AI-based evidence.
The European patent No. EP 3118356, entitled “Methods for Securely Starting and/or Stopping a Rotor of a Rotor Spinning Machine and Rotor Spinning Machine”, was filed by the Czech-based company Rieter CZ s.r.o. (“Patent Proprietor”). The patent covers a method developed to ensure safe operating conditions in rotor spinning machines, as well as a machine implementing this method. The invention focuses in particular on the control of active magnetic bearings and the prevention of automatic rotor start-up or the execution of a safe shutdown if the desired values or conditions are not met. These features enable the system to take automatic preventive action in the event of a malfunction or when safety conditions are not satisfied. The core of the examination concerned whether these features, as set out in the claims, were distinguishable from solutions known in the prior art.
The application was opposed by the Germany-based Saurer Spinning Solutions GmbH & Co. KG (“Opponent”). In its opposition, the Opponent argued that the technical content of the patent was neither novel nor involved an inventive step. It asserted that rotor control systems incorporating active magnetic bearings had already been addressed by existing techniques, and that, in this respect, the patent application did not introduce any novelty.
The EPO’s Opposition Division issued an interlocutory decision, finding that, with certain amendments, the patent complied with the provisions of the European Patent Convention (“EPC”). Following this decision, the Opponent appealed.
The main issues debated during the appeal proceedings included whether the method truly possessed novelty when compared with the known art, and whether it involved an inventive step in light of the existing body of technical knowledge. The Opponent maintained that the solution proposed by the patent was merely a known or readily predictable combination, whereas the Patent Proprietor contended that its solution provided a unique response to specific technical challenges and therefore met the patentability criteria. This decision illustrates the complexity of patentability assessments in technical fields such as rotor spinning machines, and underscores the critical role played by the concept of the “person skilled in the art” in such evaluations.
IV. THE CONCEPT OF THE PERSON SKILLED IN THE ART AND THE ASSESSMENT OF WHETHER ARTIFICIAL INTELLIGENCE CAN QUALIFY AS SUCH
Decision T 1193/23 contains significant statements regarding the role of AI-based chatbots in patent law proceedings. During the oral hearings in the appeal process, the Patent Proprietor referred to answers generated by ChatGPT to explain the meanings of the terms “position control” (Lageregelung) and the distinction between “checking” (überprüfen) and “monitoring” (überwachen)9. This raised the question of whether artificial intelligence could be considered the “person skilled in the art.” However, the Board of Appeal expressly stated that responses obtained from ChatGPT do not have direct evidentiary value for the purposes of legal validity.
The Board emphasized that ChatGPT’s responses were insufficient to accept that they accurately reflected the understanding of the relevant skilled person in the interpretation of claims. The main reasons for this, as explained in the decision, are as follows:
(i) Unknown Training Data: According to the decision, chatbot responses are based on training data that is unknown to, and non-transparent for, the user. As the scope and accuracy of this data cannot be fully ascertained, it is impossible to accept such outputs as evidence in a legal proceeding. For evidence to be legally admissible, it must be transparent and verifiable.
(ii) Context Sensitivity: Chatbot responses are sensitive to context and to the exact formulation of questions. Different phrasings of the same question, or posing the question in different contexts, can yield different answers. This undermines the reliability and consistency of such responses.
(iii) Incompatibility with the Understanding of the Skilled Person: In patent law, claim interpretation is based on the contemporaneous understanding of the relevant person skilled in the art. There is no guarantee that chatbots accurately mirror such understanding. Moreover, as they do not possess the capacity to “understand” or “think,” but rather generate text based solely on statistical models, this fundamental cognitive distinction constitutes one of the reasons why AI cannot assume the role of the skilled person.
(iv) Insufficient Evidence: The Board stated that the manner in which certain terms in a patent claim or application would be interpreted by the skilled person must be supported by tangible evidence, such as appropriate technical literature. It stressed that ChatGPT outputs do not constitute such tangible evidence.
In light of the general principles of patent law, the decision concluded that AI systems are not, at present, recognized as the “person skilled in the art.” AI may serve as an auxiliary tool, facilitating access to information and performing data analysis, but is not considered sufficient for the final assessment of patentability criteria. This is particularly evident in evaluating inventive step, a criterion inherently open to interpretation.
In T 1193/23, the Board acknowledged that AI systems can be useful in accessing technical knowledge and in preliminary analytical processes. Nevertheless, it held that the decisive standard in interpreting patent claims remains the understanding of the relevant person skilled in the art. Accordingly, AI systems are not regarded as direct instruments for legal assessment. In this respect, the decision affirms that AI is to remain a complement, rather than a substitute, for the concept of the technical skilled person.
These debates require a multi-layered evaluation at both cognitive and normative levels. Due to their training on extensive datasets, AI systems have high-capacity access to technical literature; they are free from certain cognitive limitations frequently encountered by skilled persons, such as distraction, bias, and limited memory; and their systematic and comprehensive search capabilities reduce the risk of oversight. Particularly in areas such as patent documentation and technical data analysis, AI can deliver a performance in terms of speed and scope far exceeding that of human technical experts.
Nevertheless, in patent law, the concept of the person skilled in the art is grounded not only in the capacity to access information, but also in the ability to foresee how that information would be assessed, as of a specific date, against the average level of knowledge in the relevant technical field. In this respect, the person skilled in the art is a legal fiction, serving the principles of stability, predictability, and objectivity in the decision-making process. By contrast, AI systems, owing to the opacity of their training data and the context-sensitivity of their outputs, may interpret the same technical term inconsistently under different conditions. Furthermore, the “black box” nature of their decision-making processes presents challenges for legal oversight and accountability mechanisms. The interpretation of patent claims is not merely a technically informed exercise, but also one that involves the process of construing and ascribing meaning within a defined legal framework. In light of current developments, this constitutes an area of assessment that AI is not yet capable of directly undertaking.
From this perspective, while it is acknowledged that AI systems have the potential to deliver high efficiency and offer supportive functions in technical data analysis and preliminary examination processes, it is evident under the current technical and legal framework that they lack the structural capacity to assume the role of the “person skilled in the art.” Decision T 1193/23 thus takes a clear stance against substituting AI for the technical skilled person, delineating the legal boundaries on this matter. At the same time, the approach set forth in the decision establishes a normative basis for considering the extent to which AI technologies may, in the future, be integrated into patent examination processes, particularly in light of the principles of interpretability, transparency, and verifiability, and may guide the development of new legal frameworks accordingly.
In parallel, the EPO is also taking practical steps that reinforce the view that AI systems do not replace the role of the actual skilled person in patent proceedings. In an announcement published in April 2025, the EPO introduced a pilot program, effective from May 2025, for the use of AI tools in the preparation of hearing minutes in oral proceedings conducted via videoconference10. Under this scheme, audio recordings of oral proceedings are made, minutes are prepared on the basis of AI-assisted transcripts, and the recordings are deleted after the minutes are sent to the parties. The final responsibility, however, remains with the human examiner conducting the review11.
Moreover, in the same period, the EPO launched an AI-based chatbot called the “legal interactive platform” for MyEPO users12. This platform responds to user queries with information drawn from official legal texts such as the EPC, the PCT, and decisions of the Boards of Appeal; it provides links to source materials; and it allows users to select between texts and case law13. Both initiatives highlight the potential of AI to accelerate processes and facilitate access to information, while underscoring the continuing importance of human oversight, responsibility, and supervisory mechanisms. Consequently, the framework in which AI functions as a complement, rather than a replacement, for the person skilled in the art is reinforced.
V. CONCLUSION
The decision of the EPO Board of Appeal dated 15.04.2025, T 1193/23, constitutes an important precedent both in terms of the interpretation of the concept of the person skilled in the art and the evidentiary value of AI-assisted tools in legal proceedings. While the Board acknowledged that large language models such as ChatGPT can facilitate access to technical knowledge, it made clear that, due to their context sensitivity, lack of transparency, and absence of expert judgment, such systems cannot be accepted as a direct point of reference in the interpretation of patent claims.
The decision goes beyond a purely technical assessment, delineating the role of artificial intelligence in patent law and establishing that such tools should be regarded in a supportive rather than determinative capacity. In this respect, T 1193/23 offers an authoritative evaluation for both patent law practitioners and for the broader discussion on the use of AI-based systems in legal contexts.
B. KEY TAKEAWAYS
(1)The interpretation of patent claims must still be carried out according to the hypothetical knowledge level and perspective of the “person skilled in the relevant technical field.”
(2)Responses obtained from AI-based tools such as ChatGPT are not accepted as binding evidence in the context of patent law.
(3)The non-transparent nature of the datasets on which artificial intelligences are trained reduces confidence in their outputs and limits their evidentiary value.
(4)Because chatbot responses are highly dependent on the formulation and context of the question, they are unreliable for providing a fixed interpretation of technical terms.
(5)As interpretations must be transparent and verifiable, the generative yet non-traceable nature of AI systems is restricted in legal proceedings.
(6)Decision T 1193/23 serves as an example that clarifies the rules and limitations regarding the positioning of AI within the context of patent law.
(7)AI systems deliver high efficiency in terms of time and scope in data analysis, technical document review, and preliminary examination stages.
(8)Since the person skilled in the art is defined not merely as someone who can access information, but as a hypothetical construct capable of interpreting such information within the framework of the technical knowledge level at a given date, it is presently not feasible for AI to assume this role.
(9)Decision T 1193/23 concretely sets out the cognitive, methodological, and legal criteria that must be considered when integrating AI technologies into patent law.
(10)Outputs from AI-based systems cannot replace human expert judgment in assessments that require interpretation within a normative and legal context.
