The inventor designated in a European patent has to be a human being and not a machine. That is the core message of two recent EPO refusals of patent applications in which an AI system was designated as the inventor.

The EPO decisions, concerning the applications EP 18 275 163 and EP 18 275 174, both filed in autumn 2018, were published earlier this week. Oral proceedings were held in November 2019.

As the EPO reported in a press release: ‘In both applications a machine called “DABUS”, which is described as “a type of connectionist artificial intelligence”, is named as the inventor. The applicant stated that he had acquired the right to the European patent from the inventor by being its successor in title, arguing that as the machine’s owner, he was assigned any intellectual property rights created by this machine.’

In its decisions, the EPO stated: ‘The application designates a machine as the inventor and therefore does not meet the formal requirements under the EPC (Article 81, Rule 19(1) EPC).’ And: ‘Names given to things may not be equated with names of natural persons. Names given to natural persons, whether composed of a given name and a family name or mononymous, serve not only the function of identifying them but enable them to exercise their rights and form part of their personality. Things have no rights which a name would allow them to exercise.’

The EPO also stated that the understanding of the term inventor as referring to a natural person appears to be an internationally applicable standard, and that various national courts have issued decisions to this effect.

The refusal can be appealed within two months at the EPO Boards of Appeal. The full decisions are available here:



To make sure you do not miss out on regular updates from the Kluwer Patent Blog, please subscribe here.

Kluwer IP Law

The 2022 Future Ready Lawyer survey showed that 79% of lawyers think that the importance of legal technology will increase for next year. With Kluwer IP Law you can navigate the increasingly global practice of IP law with specialized, local and cross-border information and tools from every preferred location. Are you, as an IP professional, ready for the future?

Learn how Kluwer IP Law can support you.

Kluwer IP Law
This page as PDF


  1. This must be fake news. I know for sure that U.S. patent 583,785 was filed by James Brown, who, as everybody knows, was a sex machine.

  2. How about Rule 19(2) – The EPO shall not verify the accuracy of the designation of the inventor? What would happen if you simply entered “none”?

    1. Article 90(3) in conjunction with Rule 57 f) EPC obliges the Receiving Section to examine whether the designation of the inventor has been made in accordance with Rule 19(1) EPC. This presupposes that the required indications have been given, otherwise the application has to be refused pursuant to Art. 90(5) EPC. Without this requirement, the EPO could not safeguard the inventor’s moral right to be mentioned in the patent in accordance with Art. 4ter of the Paris Convention.

  3. In answer to Light Blue: No idea. Has anybody tried writing “none”?.

    But why not? Where’s the harm? Why should the EPO object? In whose interests?

    If Applicant is not entitled, and the true owner wants to dispute ownership of the patent rights, the statement “none” renders clear for the true owner how best to plead.

  4. A few years agon there was a development for combinatorial chemistry were combinations of elements would automatically be created and analyzed for particular properties (one was fluorescence, I believe). Were an improved compound to pop out of this machine, who would have been the inventor? The fact that these two applications are in the technical field which is currently the obsession of the EPO (fifth industrial revolution etc.) does not mean that this issue is new.

    1. Who would have been the inventor might not be the correct question. Rather, why should it be an invention?

  5. Paragraph 23 is interesting “The EPC does not provide for non-persons, i.e. neither natural not legal persons, as applicant, inventor or in any other role in the patent grant proceedings”.

    Where have I heard that before? Who is a non-person?

    The decisions seem to be a triumph of form over philosophy, and it is to be hoped there are appeals.

    As for the reference in paragraph 26 to Article 60(1) – what about Article 60(2) …If two or more persons have made an invention independently of each other, the right to a European patent therefor shall belong to the person whose European patent application has the earliest date of filing, provided that this first application has been published. So in the present case if someone had filed for the identical invention a day after the present applications, they could not get a patent because the earlier application would be 54(3) prior art – whereas the earlier applicant could not get a patent because the “inventor” was a non-person.

    Given the increasing use of AI tools, I can see the possibility of infringers in litigation seeking to show that the person named as inventor was not an inventor and the inventor was in fact a “non-person”.

    Very messy.

    1. Excellent comment – I can see AI engines developing some type of legal personality at some point in the future to regulate their behaviour (which is how corporate personhood evolved)

  6. All the comments above are quite interesting but, when looking at the nitty gritty details in both applications, reality sheds a quite different view on the topic.

    The two applications underlying the decisions of the Receiving Section are relative to
    – for the first one (EP3564144) to a food container
    – for the second (EP3563896) to devices and methods for attracting increased attention, in effect a light beacon that emits very specific signals.

    Irrespective of the discussion about inventorship, it does not seem that the machine was as “smart” as it looks at a glance, since the search carried out by the EPO revealed, in both cases, very relevant material.

    In the case of the can, it is known to connect cans by their external profile. The only difference is that in the case of the application, the connecting surface is a fractal surface. It remains to be seen whether this is inventive. As the application was refused by the Receiving Section, we may never know.

    As far as the light beacon is concerned, the whole invention seems to be based entirely on studies by the applicant himself. I would say that if only the theory on which the applicant bases his application is proven, we could start to believe what is going on. It would be interesting for the applicant to provide more than a “paper” invention and to show a real device operating according to the claimed invention. In my view, this invention is close to a substantial lack of sufficiency. A search was nevertheless possible for what was understandable, which is not a lot. As the application was refused by the receiving section, we may never know.

    What is also striking is that in both cases, the notion of fractals appears. I do not think it is innocent.

    Reading the explanations given on how the invention would have been created, it is difficult to follow that on the one hand “the machine was not formed on particular data relevant to the present invention”, but on the other hand, a few lines earlier, it is said that the machine was indeed programmed to result in the said inventions. Either one or the other, but not both at the same time. The whole thing is not very serious.

    A quick glance at the references which would explain the functioning of DABUS, shows that it is referred to at least US 5659666, which never crossed the Atlantic and US 7454388 did not give rise to a grant because of problems of extension of subject matter [Art 123(2)]. For the European Patent application corresponding to US 2015/0379394, a summons for an OP has been sent. Big problems with Art 84 (or even Art 83) do not seem to augur well for the outcome of this application.

    In order to obtain a patent, it will at least be necessary to indicate the input data that allow the network to learn, as well as the correlation rules. Without these two elements there is no sufficient description. If the learning data or the correlation rules change, the output result changes. It is hard to see how all this can be described as intelligent.

    There is nothing intelligent in these kind of machines. They are only doing what they are told and if some self-perturbation of connection weights between neurons, like alleged in DABUS, should all bring the desired result, this needs a bit more explanation.

    If the applicant wanted to draw the attention on himself, he will have succeeded. But he brings nothing `tangible` to the world of patents other than a discussion that does not go very far.

    Artificial intelligence does seem to be very fashionable. I would even speak from a hype, but this hype will like other ones (biotech, big data) die as all its predecessors did. That there may be practical applications of AI is quite certain. But to translate into patents, more than some vague indications seem necessary.

    Techrights and zoobab: FINGERS OFF!! Directly or indirectly! I do not publish comments for those to be misused!!

    1. Attentive Observer, thank for the information on the applications rejected by the EPO. And i agree with you that there is a hype about IA in general and as to the patent sphere, this has generated a bubble judging from the recent growth of filings in China and the US. As to Europe, there is the mantra spread by some circles that Europe is lagging behind, so as to exert pressure in favour of relaxing the rules. You are right to put the emphasis on sufficiency of disclosure in respect of training datasets, this is a key issue. This is not an issue which is given scant attention by the EPO, as clear from their recent guidelines on IA. Another issue is plausibility (mentioned in the resolution adopted by AIPPI at the 2019 London congress).

      1. As soon there is public money to grab, a lot of people want to have a piece of the pie. The best way to grab a piece of the pie, is to frighten politicians. They want to be re-elected and do not want to be blamed to have missed an opportunity and get their country trailing behind. Hence the willingness to fork out money, even if there are doubts about the future of the topic.

        AI is another example of what lobbying can achieve, and lobbying can be very powerful, even in IP. Just have a look at the UPC…. In the latter case politicians seem to have been literally brainwashed.

        Even if the EPO Guidelines on AI are not very developed, there is one sure thing. Any attempt to assess such a patent against competitors is doomed to fail if it turns out that sufficiency is not given. Plausibility might not been sufficient either. What is needed are hard facts linking the training data to the output data.

        Without proper divulgation of the computational model and the training data, the patent will not be worth much. Also, it might be quite difficult to prove infringement. Getting a patent might thus not be the guarantee of commercial success.

        Techrights and zoobab: FINGERS OFF!! Directly or indirectly! I do not publish comments for those to be misused!!

        1. Still another issue is the need for human input in AI learning processes. See e.g. “AI and crowdsourcing are a perfect match” in

          In EPO parlance, reliance on human intervention (in the case of crowdsourcing, this actually means large numbers of human beings) in a computer process is labelled “broken chain”. There can be no “sufficient” disclosure for cognitive, inherently subjective inputs.

Comments are closed.