top of page

The Case for the Patent Protection of Artificial Intelligence Inventions

European Patent Office (EPO) legislation states that non-technical software is not patentable ‘as such’ although this is not echoed by EPO case law. Following AI advancements the case for making patenting software more accessible has never been clearer, yet the law remains unpredictable and ambiguous. This article lays out the advantages to seeking a software patent on the current scene of current technological advancement.

What are patents?

Patents are legal rights to inventions granted by the relevant body in the region where they have been filed. A British patent is granted by the UK Intellectual Property Office (UKIPO) or by the EPO. The EPO is a regional organisation containing EU countries and other European countries such as the UK and Switzerland. A patent in the U.S. is granted by the United States Patent and Trademark Office (USPTO) and a patent in Japan is granted by the Japanese Patent Office (JPO)

Although these offices have different commercial backdrops and sometimes have vastly different patent laws, the basic understanding of a patent is similar. According to the UKIPO, a patent must be new to the world, inventive and something which can be made and used. The USPTO and EPO, despite possessing distinct patent laws, offer broadly similar definitions.

In these patent regimes, a patent allows its owner to take legal action against those who use their invention without permission in return for full disclosure of the invention, which is officially published by the patent office. In the United Kingdom, patents must be renewed every year up to a maximum of 20 years. This maximum length can also be seen in other countries such as the U.S. Importantly, inventions in these geographical areas are treated differently and different exclusions as to what may be patented apply. For instance, while varieties of plants are excluded from the patent regime of the UKIPO, they are classed as a type of patent by the USPTO.

Can you actually patent software such as AI?

Despite some minor differences in patent law, the UKIPO and EPO both agree that computer programmes are not patentable ‘as such', and the same applies to mathematical methods: this is a stance echoed similarly by the USPTO. EPO case law over the last few years, however, has determined that computer software is patentable if it provides an inventive technical solution to a technical problem. For example, at the EPO software is patentable if it improves the basic operations of a computer or if the software has a technical application. However some software processes without obvious technical external effects might not be patentable. Moreover, these new software functions are often difficult to identify as inventions distinct from the normal running of a computer. The law is therefore not clear cut on whether software is patentable as, according to the EPO the technicality is not inherent to the software but actually depends on how it is used.

The use of case law to allow technical patents is similar within the UK. However, whilst the UKIPO is obliged to follow the same principles of patent law as the EPO, its case law and examination standards have created tougher regulation on software patenting in practice. As AI can be viewed as self-learning mathematical algorithms and implemented in software the case for patenting AI is particularly unclear. Moreover, the tough UKIPO stance and general ambiguity on patenting software (and therefore AI) in the UK together pose a challenge to Sunak’s goal to transform the UK into the centre of Artificial Intelligence regulation. Without changes to this patent regime, Sunak’s hard push at Washington for the UK to become a centre of AI regulation and activity remains a pipe dream, even if the UK officially takes a pro-innovation stance on AI.

The ambiguity surrounding the patentability of software is not unique to the UK, countries such as the US and New Zealand have similar restrictions. For example, despite the apparent global ‘un-patentability’ of software, 62% of all patents issued by the USPTO are software-related inventions according to Rapacke Law Group. Overall the legislation on software patents are certainly unpredictable.

Thus the answer to the question of whether software is patentable is sometimes framed as ‘yes but no.’

Is a patent the right way to protect software innovations such as AI?

The question remains as to whether patenting software is a commercially smart move for an inventor or valuable for the technology landscape. Although some consider patents a threat to innovation, there are reasons why seeking patent protection is worthwhile. There are four key reasons for encouraging patent protection for AI - increasing transparency, increasing detectability, organisational value-add and the incentive to innovate.

The first of these reasons is to increase the transparency and therefore apparent trustworthiness of AI. Amidst the increasing discussion about possible far-reaching negative effects of Artificial Intelligence it seems that patent protection could make AI seem more trustworthy. As Haseltine Lake Kempner LLP (HLK) states- "protection in return for public disclosure of an invention is at the heart of the patent system." It is sometimes commercially advantageous not to disclose the use of otherwise hidden software in a patent publication. However, transparency could help increase public confidence in the use of AI whilst helping inventors retain commercial advantage.

Secondly, as the detectability of software increases and explainable AI is increasingly required by society, hiding software will become less possible and the functionality of using copyright or trade secrets as methods of IP protection will decline.

Thirdly, patents are an excellent way of evidencing careful use of intellectual property in a business and making it more attractive to investors and shareholders alike. HLK notes that AI patents are a massive value-add for the organisation and that there is a positive correlation between firms owning intellectual property and having a higher revenue.

Finally, excluding AI from patent regimes will disincentivise innovation, as it will reduce the available published data we can work from. Encouraging patent protection for AI will mean that later AI models can build from patented inventions to more easily correct biases that will arise when collecting data from an ‘imperfect world.’ In a medical use case, for example, AI will need to be fit to help medically underrepresented populations. Without publication of such developments the use of AI will be less standardised, less easy to regulate and overall less fit for purpose.

Although patenting AI is, as Forbes states, ‘rough’ this does not entail that AI is un-patentable, or that it should not be patented. The case for more available software patents is clear through the benefits this could bring- such as transparency, confidence, protection, regulation and standardisation.


bottom of page