27/09/2023
With the advent of ChatGPT, talk has turned to whether or not the bots are coming for our jobs. As industries become more reliant upon and integrated with artificial intelligence, the intellectual property industry finds itself at a technological juncture. We must be mindful and consider the implications of AI technologies, such as ChatGPT, on the patent system. This raises the query for inventors and patent attorneys alike: “should we be using ChatGPT to draft patents?” In this article, I aim to explore some of the key considerations when navigating the patent prosecution process at the dawn of ChatGPT.
ChatGPT is a large language model (LLM) launched by OpenAI in November 2022, based on the GPT (Generative Pre-trained Transformer) architecture. The model is trained using a large quantity of publicly available text data (such as that found in books, articles and websites). Like all LLMs, ChatGPT learns to predict the next word in a given sentence through developing an understanding of language patterns, grammar and contextual relationships.
The obvious place to start for this article was to consult the model itself:
Isabel: “Should I use ChatGPT to write my patent application?”
ChatGPT: “While ChatGPT can be a helpful tool for generating ideas and providing general information, it is not recommended to solely rely on it for writing a patent application. Writing a patent application requires a comprehensive understanding of intellectual property law, technical expertise in the specific field of the invention, and attention to detail.
ChatGPT may not have access to the most up-to-date legal requirements or industry-specific knowledge beyond its September 2021 knowledge cutoff. Patent applications need to be precise, accurate, and tailored to the specific invention, which may require the expertise of a patent attorney or a patent agent who can understand the nuances of the invention and properly draft the application.”
The response highlights a limitation in the LLM’s capabilities worth consideration. Namely, at the time of writing (July 2023), the current version of ChatGPT has been developed using data valid up until September 2021. Therefore, it cannot be relied upon provide responses that encapsulate any developments in legal requirements or industry-specific knowledge beyond September 2021. If an inventor uses ChatGPT to draft a patent application, it would require further legal input from a patent attorney to ensure the information is accurate and in line with the Law. Furthermore, a patent professional will anticipate potential issues the application may face during prosecution (potentially in respect of current advancements in the field) and draft the application with these in mind. This level of nuance does not appear to be possible with the knowledge limitations placed on ChatGPT.
Another pitfall of using ChatGPT in the drafting process becomes obvious when one considers the requirements for an invention to be patentable in the UK: amongst other requirements, the claims must be new (novel) and demonstrate an inventive step. Novelty and inventive step are assessed by identifying differences between the invention and the state of the art. The UKIPO defines the state of the art as “all matter (whether a product, a process, information about either, or anything else) which has at any time before the priority date of that invention been made available to the public”. As mentioned above, it is through use of information “available to the public” that the model is trained. Therefore, it appears ChatGPT is inherently incapable of drafting new and non-obvious claims as it only has the functionality to produce text based on known knowledge (the antithesis of a novel, inventive concept).
The key message to inventors who want to seek patent protection for their invention is not to publicly disclose it before obtaining a filing or priority date. In this regard, there have been concerns raised over Open AI’s privacy policy and the ramifications for patentees relating to this.
OpenAI’s privacy policy (updated 23 June 2023) states that Open AI may use personal information for a number of purposes, such as developing new programs and services, which could amount to a public disclosure.
Accordingly, there is a high risk that using ChatGPT to draft your patent application (and therefore disclosing your invention via ‘personal information’) could count as a public disclosure and hence render any subsequently filed patent application as unpatentable, because the requirements for novelty are not met.
Furthermore, even though the current version of ChatGPT is trained on knowledge up to September 2021, newer versions will be trained using updated data.
It is important for inventors and patent professionals to consider that data relating to an invention disclosed to ChatGPT via ‘personal information’, could be used to train the model and eventually be provided as an output to another user. This risk is palpable and accordingly, many tech giants have banned employees from using ChatGPT amidst fears that usage of the chatbot could result in the leak of confidential information.
This article has merely scratched the surface of a very interesting and quickly evolving technology. Therefore, with consideration to the above, whilst inventors should use ChatGPT to generate ideas where legal knowledge is not required, they must also heed ChatGPT’s own recommendation that: “consulting with a patent attorney or agent will help ensure that your patent application is comprehensive, well-crafted, and meets the legal requirements for obtaining patent protection.”
This article is for general information only. Its content is not a statement of the law on any subject and does not constitute advice. Please contact Reddie & Grose LLP for advice before taking any action in reliance on it.
