Whether and when to change UK patent law to increase protection for AI

This article follows the recent article ‘Call for industry views on AI and IP and ViCo


In September 2020, the UKIPO launched a “Call for views on AI and IP” (see https://www.gov.uk/government/consultations/artificial-intelligence-and-intellectual-property-call-for-views ), noting the UK government’s desire to understand the implications of artificial intelligence (AI) for intellectual property (IP) policy and their desire to ensure that the UK continues to be one of the world’s best IP environments. The UKIPO posed a set of questions relating to AI patenting and AI inventorship, to explore how the UK Patents Act 1977 meets the challenges of the new technologies and whether there is a need for change. They also looked at other IP rights.

CIPA’s reply to the patent section of this “call for views” can be downloaded in full here.

Key points summarised

AA Thornton patentability expert Mike Jennings provided the following summary of the discussions within CIPA’s Computer Technology Committee that led to their AI patenting recommendations to the UK Intellectual Property Office (UKIPO):

Our consultation with industry found that many companies wish legislators to take a cautious approach when considering changes of law to protect AI – they stressed the importance of consistency and predictability, the need for a full consideration of the legal and economic impacts of any potential change of legislation, and the need to hear from a wide range of stakeholders. We agree with this. Many companies told us that their technical experts consider autonomous inventing by AI systems to be an issue for future years, and that all current inventions involve a significant human contribution. However, some AI-innovators are already reporting that their AI systems have identified or generated new solutions to complex technical problems – i.e. we are already hearing about outputs from AI systems that would have met the requirements for patentability if invented by a human inventor, but which appear unpatentable under current UK national law. This requires us to answer the question of whether current UK legislation gives an appropriate level of protection for AI innovations, or should be changed.

We concluded that there are two main areas that can be improved, and we offered the following suggestions:

(i) a statutory modification of the definition of “inventor” to better accommodate the increasing use of AI technology (see our answer to Patents Question 3 in the attached document)

(ii) a change in practice regarding the patentability of core AI technology to incentivise the UK development of such technology (see our answer to Patents Question 7).

We concluded that an amendment to section 7(3) of the UK Patents Act 1977 is desirable to clarify ownership of inventions made using AI systems. However, the first of the above two changes would also demonstrate the UK government’s commitment “to make the UK a global centre for AI and data-driven innovation”. The US government has a similar objective for supporting US industry but there is currently more scope for flexible interpretation of US law by the USPTO, and I suggest that a legislative change is needed in the UK if the UK government wishes to avoid the UK falling behind the US in terms of which AI innovations are patentable.

Specifically, current section 7(3) of the UK Patents Act 1977 refers to the inventor as the “actual deviser of the invention”, and a narrow interpretation of those words would exclude some inventions from patent protection in the UK if the human contribution when working with an AI system falls short of the current understanding of the “actual deviser”. In contrast, many of these inventions would be patentable in the US, where it is accepted that various types of human involvement are sufficient to satisfy current US patentability requirements (see the USPTO’s report on its own consultation on AI and IP published in October 2020). Therefore, a gap is appearing between the inventions which are patentable in the UK and the US, because of the specific wording of UK patent legislation that was written before legislators started to think about inventions made using AI combined with UKIPO interpretation of the exclusions from patentability set out in Section 1(2) of the UK Patents Act.

The second of the above suggestions was made because we believe too many patent applications for AI innovations are currently rejected by the UKIPO based on a strict interpretation of the exclusions from patent protection of section 1(2) of the UK Patents Act – specifically rejecting inventions made by computer scientists concerning the internal structure or operation of an AI system.

AA Thornton comments

Some concerns were expressed that a relaxation of UKIPO practice in this area might introduce an inconsistency with European Patent Office (EPO) practice, but we need to keep in mind that the EPO already made good progress in 2018 and 2019 in defining a set of “specific technical implementations” of AI which are patentable (notably when the design of an AI algorithm takes account of the technical capabilities or constraints of the system on which it will run) as well as a long list of “specific technical applications” of AI which are patentable if the patent application claims are limited to the specific technical purpose. The UKIPO needs to catch up. Furthermore, the EPO is working on a set of examples of patentable AI-related claims (similar to the ones published by the USPTO in January 2019, and examples published by the JPO in 2017) and will be publishing these with updated examination guidelines in early 2021.

We suggest that UK legislators and the UKIPO should pay careful attention to these EPO developments in early 2021, and should work with the EPO and other major patent offices to try to achieve substantive patent law harmonisation for AI technologies, but the UKIPO and UK legislators should not sit back and wait for a widening of the gap between what is patentable at the UKIPO and USPTO before taking action.

We also suggest that harmonised treatment of AI patenting is achievable, because all of the world’s patent offices are facing the same issues at the same time (note the conversation on AI and IP led by WIPO, and the USPTO’s consultation on AI and IP that preceded the UKIPO consultation). Also, although we were a long way away from harmonised international law on patentability of computer-implemented inventions for many years, we have moved much closer to harmonised treatment of many types of computer-implemented invention in recent years (e.g. including computer simulation, where the USPTO examples of patent eligible subject matter published in January 2019 were extremely close to EPO practice).

We are therefore keeping our fingers crossed for significant progress in 2021, starting with the EPO’s pre-publication of their draft examination Guidelines and examples of patentable AI claims (both expected in February), the awaiting decision of the EPO Enlarged Board of Appeal on computer simulation, and continuation of WIPO’s conversation and the conversations taking place directly between major patent offices.