Ström & Gulliksson: Use of AI-Based Tools in Patent Practice
AI-based tools are becoming increasingly prevalent in many industries, and the field of patents is no exception. A number of tools have hit the market in the last few years to help applicants and attorneys streamline the traditionally labour-intensive processes involved in applying for patents. These tools offer a number of benefits but also present certain challenges. Edmund Lobb leads Ström & Gulliksson’s working group assessing these tools and is part of a pan-European group of attorneys exploring their suitability for implementation in various areas of patent practice. Here, Ed explains Ström & Gulliksson’s current approach to the use of AI-based tools.
Pros and Cons of AI-Based Tools in Patent Practice
A significant advantage of using AI in patent-related tasks is the potential increase in efficiency and speed. For example, when searching for documents, either as part of a novelty search or freedom-to-operate analysis, AI-powered tools can generally identify documents and analyse their content in a shorter time than usually possible for patent professionals. For drafting patent applications, these tools can perform repetitive tasks (such as inserting reference numbers) more efficiently, and aid in the generation of descriptive text. In prosecution, AI-based tools can be used to generate analytical or persuasive text according to existing frameworks, such as the European problem-and-solution approach. The use of such tools can therefore reduce time and costs for delivery of certain assignments, as well as improving quality by reducing human error. This can lead to increased value for patent applications, as well as increased capacity for patent professionals to focus on higher-value tasks such as strategic planning and litigation.
On the other hand, AI-based tools currently have some significant drawbacks. The effectiveness of such tools largely depends on the quality and completeness of the data they are trained on. Poor training can lead to low-quality output, including hallucinations where an output misinterprets the content of a document or is demonstrably incorrect. Furthermore, the output of these tools often lacks the nuanced judgment and experience of human patent professionals. Patent law is complex and requires a deep understanding of legal principles, industry-specific knowledge, and strategic thinking, which is currently beyond the capability of AI-powered tools. Another concern is confidentiality, which is paramount in the patent industry. Providers of AI-based tools must be able to guarantee that any sensitive input data is not shared or located on public servers.
Our assessment
We have tested a number of tools that are powered by AI in a number of areas of our practice, including searching, drafting, and prosecution. We have found that these tools are promising in certain areas but lacking in others. The tools we have used for searching and analysing documents have been able to reduce the time spent on these tasks. In particular, identifying relevant documents in a novelty or freedom-to-operate search, summarising their content, and identifying particular relevant passages can be performed more efficiently. However, we have noticed errors in some document analysis, and always consider it necessary to verify the output of the tools with a human check.
We have tested tools for prosecution (e.g. reporting, document analysis, responding to objections), that intend to act as a co-pilot for a patent attorney, reducing the time spent on tasks without compromising their quality. However, the same issues regarding hallucination or misinterpretation of technology and prior art are also present here. Furthermore, more nuanced argumentation that is often required during patent prosecution appears to currently be beyond the capability of these tools. Again, whilst the use of these tools allows for the minimisation of redundant tasks and excess work, enabling our attorneys to focus on maintaining the intended quality, we consider it essential to always perform a human check to ensure that all outputs are thoroughly reviewed and the argumentation strategy aligns with our client’s goals, products, marketplace, and patent (and litigation) practices.
Drafting tools are flooding the market, with numerous different options already available, many of which claim to offer ‘one-click generation’ of patent applications. We have tested a number of these tools and found that, whilst they can be helpful for certain parts of the drafting process, they are not yet sufficiently developed to generate full patent applications to the required quality. Specifically, the drafting of claims, particularly independent claims, remains an area where significant improvement is needed before these tools can become truly viable. While there are certain chatbot tools available that can aid in an iterative drafting process, some expertise in using them is required to obtain a sufficiently high quality in output. At this stage, we do not consider the limited advantages offered by many such tools to be commensurate with the time and cost to implement them, although we are constantly monitoring this area for the latest developments.
Conclusion
Whilst AI-based tools offer a number of benefits in the field of patents, we exercise and advise caution when using such tools. Use of AI-based tools should currently be limited to areas where risk of hallucination is minimal, and any output should always be checked manually. The quality of any such tool should be thoroughly assessed by a patent professional before they are used in practice. Ström & Gulliksson will continue to monitor this area and maintain an open dialogue with our clients regarding our use of these tools.
Ed will attend the AIPPI World Congress in Hangzhou, October 19-22, 2024, together with Erik Broströmer. Don’t hesitate to reach out if you will also be there and would like to meet Ed or Erik!