In a recent decision T 1193/23 - 3.2.06 by the Boards of Appeal of the European Patent Office (EPO), the role of artificial intelligence, specifically ChatGPT, in interpreting patent claims was scrutinized.
The Case Background
The patent in question, EP 3118356, was opposed by Saurer Spinning Solutions GmbH & Co. KG, who argued that the patent lacked novelty and inventive step. The patent proprietor, Rieter CZ s.r.o., relied on responses from ChatGPT to support their interpretation of certain terms in the patent claims. However, the Board of Appeal dismissed the relevance of these responses, emphasizing that the understanding of a skilled person in the technical field is paramount.
The Role of ChatGPT
The patent proprietor used a Large Language Model (LLM), in the present case ChatGPT, to interpret terms like "bearing control" and "check" versus "monitor." They argued that ChatGPT's responses supported their narrow interpretation of these terms. However, the Board of Appeal noted that LLM's responses are based on unknown training data and are sensitive to the context and formulation of the questions. Therefore, they cannot necessarily reflect the understanding of a skilled person in the relevant technical field.
Key Takeaways
- Skilled Person's Understanding: The Board emphasized that the interpretation of patent claims must be based on the understanding of a skilled person in the technical field. This understanding can be supported by technical literature and expert opinions, not by AI-generated responses.
- Limitations of AI: While AI tools like Large Language Model can provide quick and seemingly authoritative responses, they lack the contextual understanding and expertise of a skilled person. Their responses can be influenced by the training data and the way questions are phrased, making them unreliable for legal and technical interpretations.
- Legal Precedent: The decision aligns with previous rulings, such as T 206/22, which also emphasized the importance of the skilled person's understanding in interpreting patent claims. In the decision T206/22, the Board of Appeal highlighted that the fact that LLM understood a claim in one understanding would not mean that the skilled man would do so. On particularly striking arguments in this respect was that the understanding of the skilled man is based upon all the information available at the date of filing, while the response of a LLM can be based on all the information available at the date the request is made.
This decision shows that the interpretation of the features of a claim must still be carried out by a human. We can hope that the same will always be true for the assessment of inventive steps, or sufficiency of disclosure, involving a "person" skilled in the art.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.
