New European Patent Office guidelines protect AI and machine learning 'inventions'

Withers & Rogers' Karl Barnfather examines the European Patent Office's 'Guidelines for Examination', which took effect on 1st November

New guidelines published by the European Patent Office (EPO) will help to smooth the way for innovators to secure patent protection for a variety of artificial intelligence (AI) and machine learning inventions, including the underlying algorithms themselves.

However, some care is still needed to ensure detailed evidence of technical effect is provided when preparing patent applications.

Before the EPO the updated guidelines, tech companies were concerned that it would take an unduly hard line

The new ‘Guidelines for Examination', which took effect on 1st November 2019, have been widely welcomed by businesses involved in the development and/or application of AI and machine learning technologies. They include a new section containing advice about how such patents should be assessed.

Despite European case law having already confirmed that the algorithms developed to run these systems may be eligible for patent protection where they make a recognisable technical contribution, there is now greater certainty for innovators.

This clarification has come at just the right time for the many hundreds of AI and machine learning inventions, which are currently undergoing examination for patent protection in Europe.

The door is open to patent applications for a wide range of AI and machine learning inventions, as long as they can demonstrate a technical contribution

The volume of patent filings at the EPO related to AI and machine learning technologies has increased, year on year, since the start of the decade, with just over a thousand AI or machine learning patent applications being filed at the EPO in 2010, increasing to just under seven thousand applications being filed in 2017. This suggests that innovation activity is still proliferating and has not yet peaked.

Before the EPO provided the updated guidelines regarding patent protection for AI and machine learning innovations, tech companies were concerned that it would take an unduly hard line. Instead, the EPO has adopted a forward-thinking stance; confirming that as long as inventions can demonstrate that they solve a technical problem with an innovative technical solution, they will be assessed and examined in the same way as other inventions.

The guidance includes a number of helpful examples, which make it clear that the door is open to patent applications for a wide range of AI and machine learning inventions, as long as they can demonstrate a technical contribution. They can achieve this either through their end use or by demonstrating that the underlying technology itself makes a technical contribution, which could include a neural network or support-vector machine for example.

A neural network could achieve patent protection in its own right, if its architecture is considered inventive and solves a technical problem

To illustrate this point, the guidelines describe how a neural network used to operate a heart-monitoring device capable of detecting an irregular heartbeat, could qualify for patent protection as it is making a technical contribution. In a different example, a neural network could achieve patent protection in its own right, if its architecture is considered inventive and solves a technical problem.

Despite the additional clarity the guidance provides regarding the patent eligibility for AI and machine learning inventions, it is still be open to interpretation. For this reason, innovators should continue to take care to provide detailed evidence of their invention's technical contribution when preparing patent applications.

Research conducted by Withers & Rogers shows that the majority of European patent applications for AI-related and machine learning inventions involve the use of existing, well-known algorithms to solve a defined technical problem. For example, 62 per cent of the European patent applications studied involved the use of existing algorithms to solve a technical problem in a field outside of AI and machine learning.

However, a growing number of patented technologies are directed to the underlying algorithms themselves. Of the recent European patent applications studied, nearly 28 per cent were related to the AI and machine learning algorithms themselves.

In summary, regardless of whether innovators are applying existing AI or machine learning algorithms to their problems to produce innovative solutions, or using their know-how to identify new algorithms which can be applied to such problems, the guidelines confirm that it should be possible to achieve patent protection for both areas of innovation.

By considering this issue and publishing new examination guidance, the EPO is signalling its interest in this exciting area of research and development activity and seeking to support innovators in commercialising their inventions. Improved clarity surrounding this important pathway to patent protection, spanning up to 44 countries, could help to attract investors and open the door to commercial opportunities and industry collaborations.

Karl Barnfather is chairman and patent attorney at European law firm, Withers & Rogers, specialising in computing and consumer electronics