Isaac Asimov famously coined the three laws of robotics in the 1940s*, but these didn’t explain whether a robot (or artificial intelligence) could be the inventor of a patent. Nor did more recent explorations in the 1980s (The Terminator) or the 2010s (Westworld and Humans). Thankfully, the Receiving Section of the European Patent Office in 2020 has sought to fill this lacuna by handing down two decisions which find that the robot (or AI system or machine) cannot be such an inventor.

The decisions were made in relation to European patent applications EP 18 275 163 and EP 18 275 174, In both cases, the applicant (Dr Stephen Thaler) indicated that the inventor had been a machine (a type of connectional AI called DABUS created by Dr Thaler) and that he had acquired the rights to the invention as employer or as a successor in title to DABUS.

The Receiving Section decided that the inventor must be a natural person; that an AI system or machine has no legal personality comparable to natural (or legal) persons; and that the application must fail on the basis of a failure to identify a natural person as inventor. The requirement that the inventor be a natural person, although not explicit, was derivable from a purposive construction of the EPC and its legislative history; from the case law of the EPO Boards of Appeal and the UK and US courts; from the patent offices in China, Japan, Korea and the US; from the patent laws of Lithuania and Estonia; and from the results of an EPO consultation of member states in 2018/19 and an AIPPI study from 2015.

This somewhat eclectic range of citations reflects the novelty of the question being considered by the Receiving Section, and some of the sources do seem to provide rather weak support for the position adopted. For instance, in the legislative history quoted in the footnotes there is heavy reliance on the use of the word “person” and even the word “his” to support the requirement that a person be “natural” (thankfully without a suggestion that this means an inventor must also be male). In the UK, the first case cited (University of Southampton, at para 39) simply uses the words “person”, “he” and “they”, while the second case (Yeda, at para 20) uses the phrase “natural person” while citing the first case with approval and without the question of “natural” or otherwise being in issue in the case (and so not being binding on the point). Finally, the AIPPI study was on a different topic (multinational inventions) and there is no mention of “natural” persons in the summary report or resolution.

None of this means that the Receiving Section is necessarily wrong to conclude that a robot (or AI system or machine) cannot be the inventor of a patent, even if it leaves some innovations unpatentable. In any case, the Receiving Section is unlikely to have the final word – the issues will be back. We can expect to see an appeal filed within the next two months (although the CrowdJustice campaign failed to meet its target), with a final decision from the EPO on these cases not until 2021 or later. ¡Hasta la vista!

* Asimov’s Three Laws of Robotics:

  1. A robot may not injure a human being or, through inaction, allow a human being to come to harm.
  2. A robot must obey the orders given it by human beings except where such orders would conflict with the First Law.
  3. A robot must protect its own existence as long as such protection does not conflict with the First or Second Laws