EPO Provides Reasoning for Rejecting Patent Applications Citing AI as Inventor

“Whether machine learning will advance to the point where a machine can truly conceive within the framework of what we consider innovative and worthy of a patent is open for debate, but we are not there yet.” – Gene Quinn

https://depositphotos.com/8265106/stock-photo-birth-of-artificial-intelligence.htmlEarlier this month, the European Patent Office (EPO) and the UK Intellectual Property Office (UKIPO) each rejected two patent applications that designated an artificial intelligence named DABUS as the inventor. While the UKIPO published a decision setting out its reasoning, the EPO simply stated at the time that the applications did “not meet the requirement of the European Patent Convention (EPC) that an inventor designated in the application has to be a human being, not a machine.” Now, the EPO has released more detail about the grounds for its decision.

The applications at issue are for a “food container” (number EP3564144/ EP 18 275 163 ) and “devices and methods for attracting enhanced attention” (number EP3563896/ EP 18 275 174). They were filed by the Artificial Inventor Project, which has filed patent applications for the inventions via the Patent Cooperation Treaty (PCT) in the United States, United Kingdom, Germany, Israel, China, Korea and Taiwan.

DABUS was developed by Dr. Stephen Thaler, who is named as the applicant on the patent documents. DABUS stands for “Device for the Autonomous Bootstrapping of Unified Sentience” and independently created the inventions at issue. “The inventions were conceived by a generative machine intelligence, judging merit of its own self-conceived ideas based upon its own cumulative experience,” according to the Artificial Inventor Project website.


In the EPO press release today, the Office explained:

The EPO considered that the interpretation of the legal framework of the European patent system leads to the conclusion that the inventor designated in a European patent must be a natural person. The Office further noted 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 press release also linked to the two nine-page decisions refusing each application, which set out the facts of the prosecution history and detailed its analysis. In both decisions, the Office explains that the applicant indicated that ‘the invention had been made by a machine and that the machine ‘identified the novelty of its own idea before a natural person did.’” The applicant further argued that “acknowledging machines as inventors would facilitate the protection of the moral rights of human inventors and allow for recognizing the work of the machine’s creators.”

However, the EPO again explained that indicating the name of a machine does not meet the requirements of Rule 19(1) EPC. That rule requires that the designation state the family name, given names and full address of the inventor and, where the applicant is not the inventor, a statement indicating the origin of the right to the EU patent, including the signature of the applicant.

“Names given to things may not be equated to names given to natural persons,” wrote the EPO. Names given to natural persons allow them to exercise rights and also “form part of their personality.” AI systems or machines have no legal personality at present, and legislation would be required to create one. T

The EPO also said that the legislative history, or Travaux préparatoires, supports the conclusion that the legislators understood an inventor to be a natural person only, as the option to add legal persons as inventors was mentioned but not included in the final draft. “In the context of inventorship reference is made only to natural persons,” wrote the EPO. Furthermore, this is an internationally accepted standard, said the decision.

In its Frequently Asked Questions, the Artificial Inventor Project explains that it does not advocate for AI to own patents, but for the AI’s owner to own patents on AI-generated inventions. Other articles on the site argue that naming a human as inventor on an invention created by an AI system “would dilute the principle that the true inventor must be identified, that is the actual deviser of the invention.”

Not There Yet

IPWatchdog CEO and Founder Gene Quinn said that the EPO’s reasoning “makes all the sense in the world.” He added:

Fundamentally, invention occurs with a conception, which grows from an idea. It is impossible for artificial intelligence to meet that basic prerequisite for what we today consider invention. Artificial intelligence is not human intelligence, but rather appears from the human perspective as if there is thoughtful intelligence.

Whether machine learning will advance to the point where a machine can truly conceive within the framework of what we consider innovative and worthy of a patent is open for debate, but we are not there yet and the laws would need to adapt to address such a future reality, should it be achieved. What seems more likely is some sui protection that would incentivize the creation of machine-realized innovation.

The decisions to refuse the two patent applications can be appealed by the applicant within two months at the EPO’s independent judiciary, the Boards of Appeal, said the EPO release.

Image Source: Deposit Photos
Image ID: 8265106
Copyright: AlienCat 


Warning & Disclaimer: The pages, articles and comments on IPWatchdog.com do not constitute legal advice, nor do they create any attorney-client relationship. The articles published express the personal opinion and views of the author as of the time of publication and should not be attributed to the author’s employer, clients or the sponsors of IPWatchdog.com.

Join the Discussion

20 comments so far.

  • [Avatar for Kuifje]
    February 24, 2020 03:03 pm

    Sorry to be late to the discussion, but I think EPO erred.

    The essence of patents is a deal: disclosure of the invention for the (perceived) benefit of society in exchange for a time-limited monopoly.

    Against that background, it doesn’t matter much whether the invention was human- or AI-generated. The invention is still there, for the benefit of mankind, even if invented by a machine, and equally worthy of protection

    The rules relating to inventorship are historical relics, aimed at (rightly) protecting the rights of the inventor, and have become part of formal requirements.

    Formal requirement should always be secondary to formal requirements.

    What the Travaux préparatoires say is not particularly relevant either. At the time it was Sci-Fi to believe in a machine-generated invention, so of course they could not even have considered such a case.

    In short: I see no credible arguments in EPO’s reasoning against allowing a machine as inventor, whereas the fundamental ‘disclosure for protection’ deal still stands.

    Now, having said that: I can’t imagine how an AI generated invention could be non-obvious…. but that is quite a different matter.

  • [Avatar for Anon]
    January 31, 2020 10:31 am


    I hear you. There are only a few things that I can recall in which we have profound disagreements.

    And yet, the degree of separation from “just use a truly mindless automaton” and the state of the art in AI does raise the questions presented herein, and does raise them now (even before we truly grasp the biological end).

    That is why I think this to be interesting — and why I keep on trying to push the discussion past the ‘just using mindless automaton’ mere machine point.

    While I personally think that any ‘reaching the Singularity’ will NOT unfold as in any number of our science fiction renditions (I think any such Singularity will be intelligent enough to not humanity be aware of its existence), so whether or not we have reached that point (of actual and true artificial intelligence in its fullest form may well be something that happens without our knowledge (compounding our lack of bio understanding), we ARE reaching a ‘disassociation’ point vis a vis a notion of ‘devisor’ and “straightforward [mere] computer-implemented inventions.”

  • [Avatar for Anon2]
    January 31, 2020 09:08 am

    @Anon 17

    So then, the insanity has already begun.

    If the technical and philosophical inadequacies of adjudicators can currently be characterized as “less than desirable”, and for straightforward computer “implemented” inventions, at that, I can only assume (or imagine to my great horror) that future attempts by those entrusted with upholding and interpreting law, to navigate through the considerations of law’s applicability to algorithms which mimic and simulate the appearance of human functions (i.e mimic consciousness, sentience, free will), will be disastrous for decades to come, if not longer.

    Centuries or millennia from now, once we have the science and actually know what in reality, consciousness, sentience, and free-will consists in, we may actually be able to create a REAL consciousness equivalent to a person, and only then things may be quite different.

    But for now, we give rights to (and impose obligations on?) blind, unthinking, automatons, to our ridicule at best and at our great peril at worst.

  • [Avatar for Anon]
    January 30, 2020 12:51 pm


    You presume a key point here in that your statement mandates an item that at least one court HAS differentiated: the UK case DOES provide that the AI IS the devisor. As a devisor, you simply CANNOT say “tools people use.”

    I “get” that IF the AI is NOT granted the status of devisor, THEN the ‘mere machine’ aspect makes ALL of this far less interesting.

    But we are no longer ‘there.’

  • [Avatar for Anon2]
    January 30, 2020 10:12 am

    Take any random law of the following type…

    Any one who… [does x]… shall be subject to fine/imprisonment/liability.
    Any person who…[does x]… shall be subject to fine/imprisonment/liability.
    Whoever…[does x]… shall be subject to fine/imprisonment/liability.

    Think of contexts in which acts constitute commission under a strict liability standard… no mens rea or “willfulness required”.

    e.g. dissemination of child pornography or making available to the public writings constituting libel

    e.g. infringement of patent rights

    Whether or not the acts x may be arguable said to have been performed by any non-person, the applicability of the law always starts with “whoever” or “one who” or “person”…

    So it matters not that it is arguable that a computer which drew or wrote something, and a photocopier which copied it and a delivery truck which delivered it, all “participated” in dissemination or publication of child porn or libel….

    or that any gadget itself is liable for making, using, or selling which constitute infringement.

    we should not be trying to fine or imprison any photocopier, delivery truck, computer, or gadget…

    Laws apply to persons or organizations recognized as persons, they do not apply to the tools people use.

  • [Avatar for Anon]
    January 30, 2020 09:40 am

    … and thanks for the link, MaxDrei.

    As I previously posted, that case does mirror a possible path in the US for copyright.

    However, that same path does NOT exist in the US Sovereign for patents.

    Even though law for both patents and copyrights derive their authority from the same Constitutional Clause, Congress has explicitly treated copyright differently in the “work for hire” sense, and as Stanford v Roche explained, the patent system in the US does NOT provide a path for juristic persons (and any non-real person) to obtain patent protection by merely being an ‘owner’ or ‘employer.’

  • [Avatar for Anon]
    January 30, 2020 09:30 am

    And yet MaxDrei, both the UK and the EPO refused to advance along that line.

    Granted, they did not do so because at least in no small part, the applicant is insisting on having the inventor reflect their view of the devisor.

    As I have noted, my path (and I am not sure that this is actually in accord with the Sovereigns of the UK and EPO cases) would have been to do as you suggest here and override the applicant’s protests, declare the applicant TO BE the inventor through the path of being the owner of the “mere machine,” and accept the applications as amended. Any challenge then to “actual inventor” could then play out without jeopardizing patent protection for the mere formalities stage (that you have elsewhere declared to be very lenient anyway).

    With protection secured (at least at the initial processing step), the ‘debate’ over inventor status could then proceed under the established steps for challenging inventorship (including exactly what parties may so challenge).

    This though is NOT the path taken in either the UK or EPO deliberations.

  • [Avatar for MaxDrei]
    January 30, 2020 06:17 am

    This court decision:


    in China is, on its face, not about patentable subject matter. Nevertheless, it is relevant to the field of machine-generated inventions. The owner of the machine is deemed to be the owner of what the machine created.

    So it’s no big deal then, the first time a machine conceives of, and reduces to practice, a patentable invention. What counts is ownership, and whether the Applicant for patent rights is or is not the owner of the patentable subject matter.

  • [Avatar for Anon]
    January 29, 2020 06:22 pm


    I do not think that an article on Treasury Stock will give you the legal context of the current discussion. That’s a whole ‘nother animal of “ownership,” along the SEC and investment control lines.

    Without a doubt, a juristic person can own a corporation. Once that is accepted, the entire line of thought in the EPO decisions becomes fallible to reasoning that places the non-real person attributes of an AI machine on the plane of a juristic person. Has this — in fact — been done? Not to my knowledge. But such does not mean that it cannot be done. The EPO rationale provided in its own reasoning includes the notion that (if consistently so chosen), the judicial actions MAY create the notion of AI as juristic person and avoid the complications that an AI lacks “personhood” or otherwise is a real person.

    I am NOT saying that such a path has been taken. I AM saying that logically (and legally) such a path may be able to be taken.

    On a different note, I do agree that much of what THIS applicant has been doing is at least partially marketing gimmick. If push comes to shove and the various courts INSIST that the AI is nothing more than a NON-divisor automaton (‘just a machine’), than what appears MUST be done is instead of outright refusal to accept (and date) the patent applications, the court need instruct the various patent offices to accept the applications, note the owner of the “just machine” AS the inventor and let subsequent actions by ANY appropriate challenger seek to change the listed inventor.

    That may be a more “proper” resolution (albeit, a less interesting one).

  • [Avatar for Ternary]
    January 29, 2020 04:45 pm

    Anon, I had a “Page has Moved” error and I am not going to re-enter my answer. It is too much of a pain in the neck. Reading the articles on the DABUS site it seems to me that the whole issue is a bit of a marketing ploy. What a waste.

    However: A legal entity can own another legal entity, I agree. A “legal entity” such as a corporation can not own itself (in the US.) I found a Wiki article that explicitly says that a company cannot own itself. https://en.wikipedia.org/wiki/Treasury_stock. Let know if you disagree.

  • [Avatar for Anon]
    January 29, 2020 01:47 pm

    I believe the discussion will make a dramatic turn if…

    Purportedly (leastwise in regards to the ‘law of the case’ for the UK case), we ARE THERE.

    The attempt to put “devisor” into a wills setting is a non sequitor and has no bearing on patent law issues.

    The takeaway (that I am imploring the likes of MaxDrei, Ternary, and even Gene to take up), is that by accepting the notion that the AI is the devisor, the statement of “cannot invent autonomously” is — and must be — false. Leastwise in the legal sense.

    There is most definitely a path here, and a relatively easy one at that: the juristic person.

    Ternary, I am not sure who you have been talking to in regards to ‘ownership’ of a juristic person, but the very nature of a corporation is so that a real person need not BE ‘the owner’ (legally). The (desired) fiction is that the trappings of ‘real person’ are an encumbrance and thus are removed for the juristic person.

    The attempted view that no government has explicitly set forth inventors as not needing to be real persons misstates the actual complexity of the issue, the fact that the issue is only now a matter of first impression, and that the EPO reasoning itself is more than a bit flawed (for example, the reasoning provides a path — related to how juristic persons are treated — that may easily [if taken consistently] provide for the item being denied. There is more than just a little “circularity” in the afforded reasoning, and anyone taking more than a passing look should be able to grasp some fundamental critical aspects — especially, in view of the non-real person attributes that already exist.

    Let’s not be in a haste to dismiss an item of true first impression (such as, for example, treating AI as “just a machine” and not capable of being a devisor).

  • [Avatar for Ternary]
    January 29, 2020 11:21 am

    A machine cannot and should not “own” anything. A legal entity (“legal person”) may be formed that owns the machine and all of its fruits. But the owner of the legal person cannot be the machine. A legal entity can be owned by another legal entity. But, as I understand it, an entity such as a corporation cannot own itself. But I may be wrong.

    Self-ownership of corporations is an interesting paradox in discussions that pop up now and then. For instance, can a corporation own itself by buying back all its shares? Overall consensus (but usually from self-declared experts) seems to be no.

    Unless a machine is awarded “person-hood status” it cannot “own” anything. A “devisor” is generally defined as “a person” in relation to wills. Bringing the machine into a “legal person” seems not to help as the machine itself is not the “legal person” but property of the “legal person.”

    I believe the discussion will make a dramatic turn if a machine shows signs of “mind.” But right now I agree with Gene that there is merely an appearance of mind/intelligence. A programmed computer is a machine, for 101 purposes, for inventorship and for ownership purposes. It cannot think abstract ideas, it cannot invent autonomously and it cannot own anything.

  • [Avatar for TFCFM]
    January 29, 2020 09:51 am

    It still seems pretty simple to me: patents are government creations, and no government has authorized award of patents for inventions not made by humans.

    The EPO’s decision appears to validate the basis of that simplicity:

    Article: “The EPO also said that the legislative history, or Travaux préparatoires, supports the conclusion that the legislators understood an inventor to be a natural person only, as the option to add legal persons as inventors was mentioned but not included in the final draft. ‘In the context of inventorship reference is made only to natural persons,’ wrote the EPO.

    Governments are free to award patents for “AI-invented” inventions, should they wish to do so, but they HAVE NOT done so (yet, anyway).

  • [Avatar for Anon]
    January 29, 2020 07:32 am

    MaxDrei @4,

    We have already distinguished the UK case and its “employer” notation as merely kicking the real question down the road but a single step and that such is not helpful.

    Please refresh yourself on the earlier exchange.

  • [Avatar for Pro Say]
    Pro Say
    January 28, 2020 06:36 pm

    . . . leaving one to wonder how the computers will react when they, too, have their computer / internet inventions unconstitutionally Section 101 / eligibility trashed by the Patent Office, the PTAB, Dist. Cts., the CAFC . . . and then denied cert by the patent-hating SCOTUS.

    Can computers cry?

    Will Congress listen to computers and restore full and unadulterated eligibility for all fields of innovation?

  • [Avatar for Anon]
    January 28, 2020 05:19 pm

    As this is a fresh thread, let me repeat a link to an interesting article published some 39 years ago (and discussing ruminations from even earlier, back to the 1950’s):


  • [Avatar for MaxDrei]
    January 28, 2020 05:15 pm

    As to ownership, under the UK Statute (for example):


    inventions made by employees are, ab initio, right from the instant of conception, the property of the employer. So, at least in the UK, it is nothing new, that the owner of an invention machine owns the creative output of the machine.

  • [Avatar for Anon]
    January 28, 2020 01:55 pm

    MaxDrei and Ternary,

    We seem to be rehashing the most latest ‘discussion’ thread on AI (without really focusing on the differences in the case of a machine being acknowledged AS “the devisor”).

    Both of your views (which no doubt may in fact be accurate) are NOT at point with the notion of a machine being taken for legal effect to BE “the devisor.”

    Besides, it is much more interesting of a legal discussion, to advance beyond a notion that the automaton is just that: a mindless automaton.

    (comments in regards to Great Apes rather miss the point that law is a sovereign vehicle for real persons and the creations of real persons – that path is more in the weeds and it is doubtful to bring much light to the current subject)

    Of much more interest is the position (not fully developed in the UK case) of OWNERSHIP of a machine providing a legal reason for chain of possession.

  • [Avatar for Ternary]
    January 28, 2020 12:53 pm

    “Artificial intelligence is not human intelligence, but rather appears from the human perspective as if there is thoughtful intelligence.” I fully agree Gene.

    It is a variation on a theme. Computers and computer-implemented inventions are machines that are created/invented by humans. They are not something that even resembles human behavior or intelligence. Many people (including judges) are deceived by the description of what a machine does, which unfortunately coincides with description of human behavior.

    It may be enlightening to read Shannon’s original 1938 master thesis on Boolean description of behavior of discrete switches. While in common language it appears that Shannon in one of his examples performs a human mathematical task (factoring of integers), in fact he teaches a machine. See: https://dspace.mit.edu/handle/1721.1/11173.

  • [Avatar for MaxDrei]
    January 28, 2020 11:33 am

    The famous quote “Cogito ergo sum” is in fact the 2nd line. The first is “Dubito ergo cogito”. For me, those two lines, taken together, throw light on the difference between human and machine “intelligence”, between a natural person and a machine made by the hand of man.

    “I doubt. Therefore I think. I think, therefore I am.”

    When machine start to experience doubts, that’s when I’ll start to concede that they are “thinking”.

    Just a thought though: Great apes think. But do they experience doubts?