AI Inventorship – Where identification of an inventor becomes a grey area, what can an organisation do to achieve the best possible outcome in securing innovation as intellectual property?
Our colleague, Coreena Brinck, was recently invited to speak on the subject of how AI is ‘reinventing’ Inventorship in the modern world at a seminar organised by the UCL Institute of Brand & Innovation Law, alongside the Chartered Institute of Patent Attorneys. During the discussions, Coreena explored a range of increasingly complex subjects and touched on where the technology is headed. (You can find a link to the full event at the end of this article)
The adoption of AI shows no signs of slowing down. According to another recent event run by the ‘Big Innovation Centre’, expenditure on funding the AI industry in the UK alone has already topped £13.8 billion across more than 2000 companies. With the UK’s Fintech sector accounting for almost a quarter of that spend, alongside considerable investment in quantum computing and the resulting patentable subject matter, the subject is expected to dominate the headlines for the foreseeable future.
Patent law often remains in a state of flux, continuously playing catch up to the changing demands of revolutionary technology. The nature of the space that Zacco operates within means that we are quite often only one step away from ideas that have the potential to change the world. Such is the nature of operating at the forefront of innovation, for the simple reason that such is the nature of innovation itself. The law often can and does adapt, but these things take time.
Even when a patent office is able to clarify grant a patent for a revolutionary new idea, a lack of global harmonisation within patents can mean that the information required to prove inventorship and patentability can fluctuate quite significantly between patent offices across the globe. Having said that, a general consensus across borders is that an inventor should be an actual person and legal conditions exist for what acts they must have performed to be named as an inventor. So, under the current tests for inventorship, the law requires a natural person (or persons) to have created (or jointly created) a patentable invention as a solution to a problem and/or to have devised the problem the invention solves.
This approach has just been confirmed by the European Patent Office (EPO) Legal Board of Appeal Decision on the two DABUS applications for which an AI system referred to as ‘DABUS’ was named by the patent applicant as the inventor. The Legal Board of Appeal recently announced (on 21st December 2021) its decision to dismiss the appeal in cases J 8/20 and J 9/20. This confirms the decisions of the Receiving Section of the EPO to refuse the applications EP 18 275 163 and EP 18 275 174, in which ‘DABUS’ was designated as inventor in the application forms.
This decision also refuses the auxiliary request according to which no person had been identified as inventor but merely a natural person was indicated to have “the right to the European Patent by virtue of being the owner and creator of” the artificial intelligence system ‘DABUS’.
So what happens then if there are no human inventors who can be named?
Under the current laws for determining inventorship, if an inventive contribution has been made by a very advanced AI system, and this has led to the invention claimed, then it may not be possible to identify a human inventor who passes the test for inventorship. This decision by the EPO means that the patent application will not be accepted when the AI system is named as the inventor.
Now this has never been an issue until recently. Computers are used by inventors as tools so the actual computer does not make an inventive contribution but as technology has evolved, so too has its role. Traditionally an inventor is the person who finds a solution to a problem which is novel and inventive over the prior art, and computers were used as tools to find such solutions. To configure a problem and/or to configure a computer to solve that problem in a particular way is a longstanding mainstay of innovation, and there will normally be a human inventor who can be named in such situations.
Very advanced AI computer systems, however, may not be configured by humans to solve a specific problem. They may be programmed more generally to tackle a broader task and, in some very advanced systems, that broad task may itself be set by another computer which can autonomously develop it. In other words, computers may be configured to have a certain degree of freedom to explore data and find solutions autonomously without a human defining the problem the invention addresses. The systems themselves may also be able to recognise a “good” solution if they find one. Computers that are capable of behaving with this degree of freedom are much more likely to be capable of making an inventive contribution or of actually inventing something. Examples of such computers already exist in the form of “Generative Adversarial Networks’ if these have the capability to find problems and their solutions autonomously.
So what does make someone an inventor?
It is more than simply finding something that already exists. So someone who identifies a scientific fact, has not “invented” that fact, they have instead discovered it.
By way of example, consider an AI system which is connected to a 3D printer. If the AI system autonomously generates real items using the 3D printer, then a passer-by who picks up the real item for the first time would not normally be considered the inventor of that real item. If that person had used the AI system as a tool to some degree such as if they were involved in the system configuration or training, or the selection of the training data, to the extent that the output of the AI system would be expected based on their involvement, they could be considered an inventor under the current legal tests for inventorship. Similarly, if the passer-by who picked up the object had a ‘eureka’ moment realising it could be inventively used for a particular purpose to solve a particular problem, then that person could also, under the current legal tests for inventorship, call themselves an inventor.
By way of another example, consider second use inventions. These are inventions which are created when someone/thing realises something already known can be used in a different inventive way. For example, someone found that aspirin could be used to treat heart issues in a lower dose than that used to treat headaches for which a “second medical use” patent was granted. However, in the same situation, imagine if an AI system was configured to read all medical literature ever published. For example, the AI system might be configured only to find ‘good uses for known medicines’. Given the huge number of medicines already known, if the AI system had come up with using a lower dose of aspirin to treat heart disease then it could be difficult to find a human to name as the inventor of the second medical use, at least under current tests for inventorship.
What does this mean if you are using very advanced AI systems and want to patent an invention where the AI system may have made an inventive contribution?
Under the current law, a patent can only be granted if a human is named as the inventor. It will be very important to make clear any concerns about AI systems acting as an inventor when the patent application is being drafted. In 99% of cases, there may be a way to claim the technology so that a human inventor can be named and the legal tests for inventorship in some countries, the UK for example, may change to make it easier to be able to name at least one human inventor when an AI system has made an inventive contribution. Even if your Patent is still in the pre-grant stage then there is still time to consider inventorship issues if these are a concern.
Why is it important to name inventors correctly?
Firstly, it affects patent ownership as the rights to the patent application in the first place belong to the inventors. Secondly, patent litigators are likely to investigate who the inventors are for any patent asserted against infringers. In some countries, such as Germany, you may even lose your standing to sue if you are not the correct owner of a patent. Similarly, some countries, such as the US may even strip you of your patent rights if it turns out that there was a deliberate attempt to conceal the true inventors.
If you are working with advanced AI and you are concerned about inventorship, we can help you to navigate the increasingly complex rules, regulations, and considerations involved when attempting to patent your innovative technology. Similarly, if you have concerns about the potential for future patent litigation due to concerns over any inventive contribution by an AI system, then we would be happy to help you explore what options might be available to you.
For more information, please feel free to reach out to Coreena Brinck, who can help you to identify the next steps on your AI innovation journey.
You can find the UCL/CIPA event where Coreena spoke here.