The preview-version of the EPO’s Guidelines for Examination is out now (link here), and, for the first time ever, there is specific guidance on patenting AI applications – Are there any surprises in the Guidelines?

Earlier this year the European Patent Office (“EPO”) held its first ever conference on patenting artificial intelligence (“AI”). Following intensive discussions on the impact of AI in the patents sector, the EPO promised to update its Guidelines for Examination, to provide specific guidance on the examination of AI applications under existing computer-implemented inventions (“CII”) practice and case law.

The EPO has delivered on its promise. A preview-version of the new Guidelines is now available on its website. These Guidelines are set to take effect on 1 November 2018. Plot spoiler alert – the new section on AI and machine learning now appears in section 3.3.1 of Part G of the Guidelines.  Those of you who are familiar with the Guidelines will immediately spot that this new section on AI and machine learning has been inserted under the part of the guidelines that deals with mathematical methods. Unsurprisingly, then, the new section begins by saying that, whatever other labels “AI” and “machine learning” might carry, they are based on computational models and algorithms, and (it appears) are to be treated as such for the purposes of examination:

“Such computational models and algorithms are per se of an abstract mathematical nature, irrespective of whether they can be “trained” based on training data.“

The Guidelines go on to say that, when examining whether the claimed subject-matter has a technical character as a whole, expressions such as “support vector machine“, “reasoning engine” or “neural network” are “looked at carefully“, because “they usually refer to abstract models devoid of technical character.” (Emphasis added.) Noted. What the EPO seems to be saying is that terms alone will not necessarily in and of themselves convey a technical function, rather they will be examined to see whether what they are describing  actually contributes something that is considered technical in nature. This isn’t a new principle, although perhaps it needs restating in the context of AI, which, some might say, can be particularly jargon-heavy.

The guidance on technical contribution in the context of AI and machine learning applications is short. However, particular attention is payed to AI or machine learning applications that classify information.

“The use of a neural network in a heart-monitoring apparatus for the purpose of identifying irregular heartbeats makes a technical contribution” (emphasis added).

However an AI that merely classifies digital images, videos, audio or speech signals based on low-level features (such as edges or pixels) will not necessarily do so. Attention is drawn to cases T 1358/09 and T 1783/6 which concerned computer programs that classified text and data respectively. In those instances, it was not established that either computer program made a technical contribution. The bottom line seems to be that even if the AI comprises a really useful classification algorithm, that alone will not convey a technical purpose. Again, this is not a new legal principle. The Guidelines conclude that if, however, a classification method serves a technical purpose, the steps of generating the training set and training the classifier may also contribute to the technical character of the invention if they support achieving that technical purpose.

So, what can we learn from all this? Well, the subtext appears to be that, whilst AI and machine learning might be a new addition to the Guidelines, unless there is a good reason to deviate, the usual rules on patenting mathematical algorithms or computer programs will apply. Patenting AI is not necessarily going to be any easier …. or harder than before. What’s interesting about this is that there is public debate about the sort of protection that should be afforded to AI applications. Whilst trade secrets and copyright might seem like the most natural forms of protection, there is an interest in maintaining transparency around the use and development of AI. A number of well-known public figures (Stephen Hawking, Elon Musk… ) have called for this to ensure that AI is used responsibly. This incentive will not be unfamiliar to the EPO; indeed, one of the discussions at the EPO’s AI conference this year centred on whether or not it should be made possible (or easier) to patent AI, precisely because it would encourage the level of transparency being called for. Filing a patent requires the applicant to disclose some information about the AI and possibly the way it is being used. The EPO hasn’t, with this latest update to the Guidelines, reformed the rules yet, but as AI becomes increasingly widespread, and its autonomous functions improved, this could be a change to come.

Whilst the updated guidelines addressing AI and machine applications might not be wholly surprising to practitioners, arguably more interesting will be the treatment of inventions that have resulted from AI or machine learning under the new Guidelines. We will be taking a more detailed look at some of the other updates to the Guidelines to see what can be gleaned.

[View source.]