top of page
shutterstock_526919722.jpg
shutterstock_1115877836.jpg

01

The seven challenges of patenting Artificial Intelligence

21st May 2024

By Yannis Skoulikaris, Director

Patenting AI-related technology is not a simple task. Patent laws were originally drafted with very different technology in mind, at a time when there was a more or less clear-cut divide between the three major technology areas of mechanics, chemistry and electronics. This immediately presents a challenge, since AI combines every area of technology, producing an interesting cocktail. Patent attorneys and patent examiners therefore have to deal with two or more areas of technology in the same application. This calls for inter-disciplinary skills and a creative approach in a traditionally conservative field of law.

The second challenge is more subtle and has to do with the make-up of AI technology and the exclusions of patentability enshrined in patent law. Conventional wisdom has it that AI progress is due to extremely powerful computers, highly sophisticated algorithms implemented in software, simulation of human learning and availability of big data. Almost all of these essential elements, with the exception of computers, are excluded from patentability when claimed for themselves, at least according to European patent law.  So, it is difficult for the inventor to explain to the patent examiner why a new learning algorithm or the use of specific big data can make the inventive contribution in a particular case.

A third challenge arises from the fact that the function of an AI system is not entirely transparent. In other words, one cannot expect that the reasoning and outcome of an AI system can be completely explained, as would be the case for a conventional, non-AI computational system. The lack of a full and traceable line of argumentation and an explainable outcome from a machine learning-related process creates difficulties, especially in those cases where the logic and reasoning have to be transparent, such as in a diagnostic system.

The fourth challenge is not a purely patent law-related one, as it has to do with the lawful use of training data for a machine learning system. If such data were used without authorisation, the data owners could be eligible to seek damages. In December 2022, the Financial Times sued OpenAI and Microsoft, accusing them of using a vast number of FT articles without permission in order to train chatbots in Open AI’s systems. Even if, strictly speaking, this case is based in copyright law, it remains to be seen whether patent law will be affected.

The fifth challenge has to do with obviousness (in patent jargon ‘inventive step’) and the notion of the skilled person, two essential elements of patent law. When judging obviousness, the patent examiner asks themselves whether a skilled person (defined as a person in the field who is knowledgeable, but does not have to use inventive skill, imagination or creativity) would have thought of the invention after having knowledge of the documents in the search report. If the examiner judges that the skilled person would have thought of the invention, then the invention is obvious and cannot be patented. However, if the skilled person uses AI as a standard tool, this changes the rules of the game. A possible consequence would be to raise the threshold of inventive step, i.e. ask for a more substantial inventive contribution.

The sixth challenge is around inventorship: should an AI system be named as inventor in a patent application? Most jurisdictions specify a human inventor, but the development of AI has stirred up a lot of discussion, especially following the well-known DABUS case, where numerous patent applications were filed across many jurisdictions, designating an AI system (DABUS) as inventor. The majority of those jurisdictions decided this was unallowable, and in certain others the case is still pending.

Finally, the seventh challenge is quite unique to the patenting world: in European patent law there is a prohibition against patenting technology that is deemed unethical. The current lively discussion about ethical use of AI, for instance regarding bias towards gender and/or age in training machine learning systems, is potentially caught by this. Moreover, regulatory bodies such as the European Commission and international organisations such as OECD and UNESCO have issued guidelines on AI that add further complexity. How examiners will treat applications that might contravene the ethics provisions of patent law remains to be seen.

For the time being, patent offices do not appear to be unduly disrupted by these developments, but as the intensive research and commercialisation of AI continues to gather pace, they are likely to surface and make the headlines at various points. We at TTN have particular expertise on AI patenting. As such, we continue to monitor closely such developments, and the practice of the patent offices, in this fascinating technology. Watch this space!

02

Project Name

This is your Project description. Provide a brief summary to help visitors understand the context and background of your work. Click on "Edit Text" or double click on the text box to start.

03

Project Name

This is your Project description. Provide a brief summary to help visitors understand the context and background of your work. Click on "Edit Text" or double click on the text box to start.

04

Project Name

This is your Project description. Provide a brief summary to help visitors understand the context and background of your work. Click on "Edit Text" or double click on the text box to start.

bottom of page