The benefits of AI technology are varied and could truly have potential to revolutionise many aspects of life, including within the world of IP, but the introduction of AI technology also poses a number of challenges within the IP industry which are likely to need to be addressed in the near future. In this article, we explore some of the potential difficulties we expect to emerge with the development of AI technology and the questions which are ultimately likely to be heard by the Courts.
There is no doubt that AI machines have the ability to create subject matter which can be protected by IP. AI machines could easily create a work of art, write a piece of literature, design an object (and perhaps even 3D print it) or develop a new brand name. It is even foreseeable that an advanced AI machine could be responsible for the development of new technologies or new pharmaceutical drugs which might attract patent protection.
As the ownership of a trade mark is not linked to the person who devised the mark, it is unlikely to have any implications on trade mark ownership if an AI machine creates your trade mark, but for other forms of IP, the ownership of AI created matter is a hot topic. The first owner of a piece of work protected by copyright is generally the ‘author’, or the person who created the work. Likewise, the first owner of a design right is the designer and the first owner of a patent is the inventor. In all of these cases ownership is directly linked to the creation of the subject matter. If it is therefore accepted that AI machines can create the subject matter, does this make the AI machine the first owner of the IP?
Allowing AI to own IP has consequences in the field of enforcement and infringement of IP. If AI has sufficient legal personality to own IP then surely it must also be able to bring infringement proceedings, be sued for infringement, enter legal contracts and in all other ways be treated as a human for legal purposes? The question is actually a much bigger legal query regarding whether AI machines are capable of being a legal entity. The world has already seen the first robot to be granted citizenship of a country when a humanoid robot named Sophia became a citizen of Saudi Arabia in 2017. It is very early days in the discussion of AI machines having rights equivalent to humans and it is unlikely that we will see legislation acknowledging the legal personality of AI machines in the near future, but deliberations of the issue are only likely to increase in the coming years.
In recent decisions of the European Patent Office (EPO) the idea that an AI machine could be named as the inventor in a patent application was dismissed. The decisions relate to two patent applications that designated AI called DABUS as the inventor. According to the minutes from the non-public hearing, one of the main lines of argument from the Applicant of the patent applications was that the European Patent Convention does not mention that the inventor must be a human. This was the first test of its kind before the EPO and was dismissed after 21 minutes of deliberation by the EPO Receiving Section and the patent applications were refused. The Applicant only tried to claim the AI machine was the inventor, not that it owned the patent application. The matter raised a number of interesting questions about rights of an AI system and how ownership had transferred from the inventor to the Applicant which were not addressed to the satisfaction of the EPO. The dismissal of the arguments put forward in this case suggests that the EPO at least is not yet ready to take the idea of AI ownership of IP rights seriously. According to the decision, the EPO boards of appeal have so far not been called upon to decide on the question on whether an entity other than a natural person can be recognised as an inventor. The decision from the EPO Receiving Section is open to appeal by the Applicant and it will be interesting to see if the decision is appealed. Similar patent applications with DABUS as the inventor were filed at the UK Intellectual Property Office (UKIPO) but those applications were not allowed by the UKIPO’s Hearing Officer who said that although DABUS created the inventions, it was a machine and not a natural person. The UK legislation refers to the inventor as being a person and this was intended to mean a human person, so naming of an AI machine as an inventor was not allowable. The Hearing Officer noted that there was no law that allows the transfer of ownership of the invention from the AI inventor to the owner in this case, as the inventor itself cannot hold property. We therefore do not yet have answers to a number of questions, most importantly, where subject matter was genuinely created by an AI machine, who is the first owner of the IP?
The flip side of whether AI can own IP is whether it can infringe it. If it is accepted that an AI machine can generate subject matter, who will be held responsible if that subject matter infringes third party IP? The question is particularly interesting in the context of copyright infringement which requires actual copying, i.e. the author of the infringing work must have had access to the work protected by copyright. In the case of an AI machine, which we must expect will have access to everything on the Internet, the hurdle of showing that the infringer had access to the protected work might be much easier to overcome. Again, the concept of AI creating subject matter raises a number of complex legal queries, does an AI machine have sufficient legal personality to be sued for infringement? If not, who is responsible for the infringement of copyright by an AI machine and is it just to hold an individual to account for the actions of the AI machine and would thin individual meet the criteria of having copied the work?
Consumers play an important part in the role of trade mark law. Contentious proceedings between trade mark owners raise a number of questions which involve looking at the matter from the perception of the average consumer. In particular, the question of whether the average consumer would be confused is often central to a trade mark dispute.
With the emergence of technologies such as Amazon Dash buttons, and home appliances which can re-order staple groceries when you run out, it is conceivable to see a world where an AI machine will order your shopping for you. We must therefore ask ourselves what impact this will have on the average consumer for the purpose of trade mark law. Is it possible for an AI machine to be confused and to order the wrong product? Could an AI machine raise the level of attention paid by the relevant consumer if it ever becomes commonplace that certain goods are ordered by AI and not selected by humans?
There is also the question in patent law on whether the concept of a skilled person will change as a result of the surge in AI technologies. The skilled person is a notional person that appears in the EPO and UK legal provisions including when considering inventive step. The skilled person is considered to know concepts and terminology in the field of the application and will have means for routine work and experimentation. Should the notional skilled person be assumed to have access to commonly-used AI tools? If this is the case, does it make it easier to find a lack of inventive step for AI-related inventions or perhaps all inventions? If new legislation is introduced which allows patents for inventions made solely by an AI system, could the skilled person then also be an AI system? These questions are yet to be answered but will likely be answered through the development of case law in the field.