The UK Supreme Court DABUS Decision: The End or Just a Bump in the Road for AI Inventors?

“The Supreme Court is not the right forum for policy change since it must necessarily implement existing laws and statutes. However, the Dr. Thaler cases…highlight the need for any ongoing legal review to keep up to date with the drastic and rapid changes in technology.”

DABUSAs reported on IPWatchdog, the UK Supreme court recently ruled that a trained neural network (an Artificial Intelligence known as DABUS) could not be listed as the inventor on two patent applications filed by Dr. Stephen Thaler at the UK Intellectual Property Office (UKIPO). As a result, the two applications were treated as withdrawn for failing to comply with the requirements of the UK Patents Act 1977.

This not a surprising decision for reasons that will be set out below, and it is consistent both with the earlier UK court decisions, and with the decisions of other jurisdictions around the world where Dr. Thaler has argued his case. The decision has, however, sparked interest in the questions of artificial intelligence and its ability both to “autonomously” devise inventions and to subsequently own them.

The Legal Background and the Supreme Court Decision in Context

When filing a patent application in the UK, it is necessary according to the requirements of the UK Patents Act 1977 to list both an applicant and an inventor. Where the applicant is different to the inventor, it is also necessary to indicate why the applicant is entitled to file the application on the inventor’s behalf.

The inventor is assumed to the person who devised the invention, and therefore the initial legal owner of the invention as defined in the application as an instance of property. Further, for the applicant there must be some means by which the rights in the invention have been transferred, either by an assignment of the rights, or by operation of law. In the UK, inventions made by an employee, in cases where devising inventions is an anticipated part of the employee’s duties, automatically belong to the employer under the provisions of the UK Patents Act.

The UKIPO is not required to check whether the inventor and applicant are entitled to be so listed, and in their formalities examination of the application do no more than check that the application forms have been completed as expected. Nevertheless, mechanisms exist under the UK Patents Act for third parties who are in fact the rightful inventor or owner of the invention in the application to challenge the applicant and inventor details, and lay claim to the application or to the resulting patent.

In this context, the UK Patents Act therefore operates on the assumption that inventions defined in patent applications are a class of property that can be owned by people or companies in the same way as other forms of property, an inventor is a natural person, and the applicant is either a natural or legal person (a company or other corporate entity).

It is therefore no surprise that Dr. Thaler’s argument that DABUS should be listed as the inventor failed, since, as noted by the Court, “Whether or not thinking machines were capable of devising inventions in 1977, it is clear to me that Parliament did not have them in mind when enacting this scheme. If patents are to be granted in respect of inventions made by machines, the 1977 Act will have to be amended”.

The legal position is clear then. But as a matter of policy, this still leaves open the question of whether the law should be amended in future.

The Consultation of the UKIPO

Whether to amend the law to account for inventions devised by artificial intelligence was considered by the UKIPO in a public consultation that ran from October 29, 2021 to  January 7, 2022. Among other questions, the consultation sought respondents’ views on “patent protection for AI-devised inventions” and specifically asked “Should we protect them, and if so, how should they be protected?”.

Four options for reply were given to respondents, including: 1) “Option 0 – no legal change” [thereby preserving the current legal position discussed above]; 2) “Option 1 – expanding the definition of inventor to allow humans responsible for the AI to be named as inventor”; 3) “Option 2 – allow AI to be named as inventor or remove requirements to name an AI inventor”; and, finally 4) “Option 3 – introduce a new patent-light right to protect inventions devised by AI”.

The consultation was launched after the initial decisions by the UKIPO, the UK High Court and the UK Court of Appeal to refuse Dr. Thaler’s applications, and the question of AI inventorship and ownership were already in the public consciousness. The results are available for review here and here.

The report noted that “We [the UKIPO] received 48 written responses that expressed a view on patents and AI inventorship. Respondents included AI developers, the pharmaceutical sector, small businesses, academics, and legal and IP professionals. Respondents had mixed views on the options proposed in the consultation. The majority preferred no unilateral change in UK law for now (Option 0).”.

Reasons for this were varied but included a sense that “there is limited technological, legal, and economic evidence for legal change at present. In particular:

  • there is no consensus on who should own patents when AI is the inventor,
  • respondents consider the technological case for change may be stronger in the future,
  • international harmonisation is seen as a prerequisite to successful change,
  • the majority took the view that a unilateral change to UK legislation would be premature and may be negative for the UK’s innovation landscape”

In the replies, there was nevertheless an acknowledgement that in future:

[R]ules will need to change to take account of the increasing contribution of AI in the research and development process. This view was a particular concern of the pharmaceutical industry. Many respondents felt that once AI reaches a stage where it could invent, any inventions devised by that AI should be patentable. But respondents from the tech sector acknowledged that AI is not yet advanced enough to invent without human intervention”.

An additional consideration from the consultation was the wider discussion around protection for works generated by AI. In this context it was noted that the UK remains one of the few countries in the world that provides 50 years of copyright protection under the Copyright Designs and Patents Act 1988 for “computer-generated works without a human author”, but that this provision remains rarely used, and consequently there are still areas where its implementation may be unclear. Although there were fears that allowing copyright protection for computer generated works would crowd out human creativity, there is no evidence to date that this has happened.

The Analogy with a Team of Human Inventors

Dr. Thaler’s assertion in the cases heard before the UK courts was that DABUS was the original deviser of the inventions listed in his patent applications, that it was therefore correct to list DABUS as the sole inventor of the inventions, and that he was subsequently entitled as the owner of DABUS to be listed as applicant.

At present, the term “artificial intelligence” is perhaps best understood as referring to a trained neural network, capable of receiving a prompt and providing an output based on information that the network has “learned”. In order to provide a useful result, human operators spend  considerable time and effort training the network, so that its output matches expected parameters. In operation, technicians then prompt the network and assess the relevance of its output. The value of the trained network is that it can operate more efficiently and reliably when given new data, spotting new patterns, and arriving at new results more quickly and even sometimes more reliably than a team of humans alone.

In this context, philosophical questions arise as to whether the artificial intelligence is merely a useful “design tool” supporting a team of human researchers, is “a skilled digital assistant” working alongside the team but not necessarily capable of making an inventive contribution in its own right, or alternatively, is an entity entirely capable of inventive activity in its own right and therefore the “actual devisor of the invention”. If an instance of AI was considered to be the “actual devisor of the invention” then (as discussed above) the law would need to change in some way to acknowledge the existence of an additional type of inventor. Commentators have pointed out that the present arrangement may lead to difficulties in identifying for legal purposes the contributions made by the AI compared to, for example, the operators or engineers working with it.

Difficulties in identifying the actual inventor is not limited solely to considerations of AI, however, and already arise to some extent in assessing ownership of inventions developed by a team of researchers or product developers. It might be the case for example that not everyone in a team that produced an invention (claimed in a patent application) should necessarily be listed as an inventor. Members of the team that did vital work, but that merely followed instructions of others, and who ultimately made no inventive contribution themselves would not tend to be listed. In many cases, an artificial intelligence assisting in research or development could be viewed in this way, in other words as a design tool, or at best non-inventive digital assistant. In this regard, AI has already (within the context of a tightly controlled human defined framework) proven invaluable in identifying candidates for new antibiotics, medicines and materials.

Further, under the present test for “inventiveness” carried out by the patent office, a claimed invention must not have been obvious to the skilled person. The skilled person in the test is a theoretical individual or team of experts who has read all of the relevant prior art literature but who is assumed to have no inventive capacity themselves. Alongside the concept of an all-knowing but ultimately uninventive human “skilled person”, how is the law to assess whether an AI has actually devised a new and “non-obvious” invention?

The terms “inventor”, “invention” and the capacity to “make an invention” are all therefore given particular meanings under the present UK Patents Act (compared with what these terms may mean in the public mind), and resolving how these questions should be answered in connection with AI assisted or even devised inventions in future will require careful consideration. Indeed, one of the concerns flagged up in the UKIPO consultation was that any changes made to UK patent law should not diverge from international patent law, or else risk having a negative effect on UK applicants wishing to file overseas.

In this case, the Supreme Court is not the right forum for policy change since it must necessarily implement existing laws and statutes. However, the Dr. Thaler cases have succeeded in bringing the matter to the public consciousness and highlight the need for any ongoing legal review to keep up to date with the drastic and rapid changes in technology.  If an AI is ultimately deemed capable of being the actual devisor of an invention and of owning that invention, the question remains – would it just be patent laws that require serious overhaul?

Image Source: Deposit Photos
Author: 72soul
Image ID: 12199196 


Warning & Disclaimer: The pages, articles and comments on 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

Join the Discussion

3 comments so far.

  • [Avatar for Anon]
    January 22, 2024 05:55 pm

    Thanks – but I was aiming at the difference in implications to patent law (there are several) with the difference between, as you put it (mind you, that’s a nice way of putting it): discriminative AI and generative AI.

    Discriminative very much resembles “just used as a tool,” while – clearly – generative’s distinction on that point carries SEVERE patenting issues.

  • [Avatar for David Lewis]
    David Lewis
    January 22, 2024 01:32 pm

    Although distinguishing between generative and discriminative AI may have added some clarity, I think it is clear that the decision was intended to cover, even generative AI. Also, I think that it is clear that the Court’s decision and UKIPO report were predominantly discussing generative AI.

    Generating something original would seem to require generative AI (as opposed to discriminative AI). Consequently, I don’t know that the Court or UKIPO needs to expressly specify that distinction, because they appear to be assuming that the “invention” is patentable, which legally must be “non-obvious.”

  • [Avatar for Anon]
    January 21, 2024 04:06 pm

    One aspect not touched upon (but in my view, should have been touched upon) is distinguishing between what WAS known as AI and what has been developed and is known as (a different thing called) Generative AI.

    The two should not be confused has being the same, nor having the same capabilities.

Varsity Sponsors

IPWatchdog Events

Patent Portfolio Management Masters™ 2024
June 24 @ 1:00 pm - June 26 @ 2:00 pm EDT
Webinar – Sponsored by LexisNexis
August 22 @ 12:00 pm - 1:00 pm EDT
Women’s IP Forum
August 26 @ 11:00 am - August 27 @ 5:00 pm EDT
IP Solutions, Services & Platforms Expo
September 9 @ 1:00 pm - September 10 @ 2:00 pm EDT

From IPWatchdog