Patenting AI-generated inventions – is patent law acting the luddite?

Not limited to the field of life sciences, obtaining patent protection for an invention generated by artificial intelligence (AI) has been impeded by patent offices around the world refusing to register AI as the inventor. This issue certainly is not new (cf. our 2021 article on Digital Health), but remains unsolved. With the rise in AI-generated inventions in the life sciences field, the need for a solution is growing.

Virtually no registration of AI as the inventor

While it is commonly accepted that inventions can occur when artificial intelligence is at work, obtaining a patent truthfully identifying AI as the “brains” behind the invention by naming it as the inventor is currently impossible in most jurisdictions. One player in the field actively challenging this granting practice is Stephen Thaler, the inventor behind artificial intelligence system DABUS (Device for the Autonomous Bootstrapping of Unified Sentience). DABUS itself is patent-protected and has come up with a number of inventions, some specifically generated – upon instruction – for the purpose of initiating the quest for formal recognition of AI as their inventor. With the exception of South Africa and to the extent known, patent offices in all jurisdictions, including the EPO (decision of 21 December 2021, case no. J 8/20 – Designation of inventor - artificial intelligence; decision of 21 December 2021, case no. J 9/20 – Designation of inventor – DABUS II) and the German Patent Office (affirmed by the Federal Patent Court, court order of 11 November 2021, case no. 11 W (pat) 5/21 – FOOD CONTAINER; court order of 21 December 2021, case no. 18 W (pat) 28/20 – DABUS) to date have refused to do so.

Why are patent offices refusing to grant patents for AI-generated inventions?

The main reason for patent offices’ refusal to register AI as the inventor of AI-generated inventions is their underlying legal framework, which is interpreted as requiring the inventor to be a human being. Arguably, neither the German Patent Act (GPA) nor the European Patent Convention (EPC), foresaw the concept of an inventor being anyone but a human being when drafted.

Why the reluctance to adapt the granting practice? For one, it seems to be a question of conviction. Although there are ample arguments for inferring from the legal framework that an inventor must be human, there is no express stipulation to this effect. Thus, patent offices could change their views in light of technical progress. However, when longstanding principles are challenged, the patent world – not unlike any other – is hesitant to change. This brings to mind the lengthy debates on the “correct” patent filing system and the concept of absolute or relative novelty, grace periods etc., which re-emerged when the US reformed its patent filing system in 2013 (and created a new, hybrid system combining both approaches), or more recent discussions in Germany preceding the codification of the requirement of proportionality in the context of an injunction.

In this case, however, conviction and longstanding legal tradition alone are not the (sole) reason for hesitation. Rather, inventorship as recognized by the EPC and the GPA (‘inventor’s honour’) does not only award personal rights to the inventor, but also, e.g., the right to file a patent application or to transfer the same to a third party. Such rights require that the person/entity awarded the same is (legally) competent to bear them. Lacking legislation conferring such competence to AI, there is currently no basis on which AI could bear and exercise such rights. Insofar, registering AI as the inventor in a patent application would leave the associated rights without a legally capable rightsholder.

Relevance to the field of life sciences

Reliance on AI to analyze large volumes of data, including medical images, has become common practice also in the field of life sciences. However, AI has also become capable of autonomous knowledge and generating novel concepts. Reported AI inventions generated in the field of life sciences extend from the area of drug discovery, e.g. the use of AI to design and synthesize new compounds, to drug design, e.g. designing a target molecule, and drug repurposing, e.g. identifying new uses for existing drugs. Insofar, the ongoing discussion on AI inventorship is highly relevant also for life sciences.

Concerns such as over-compensation of the person named as the inventor instead of the AI aside, it does not seem fitting that the legal field most attuned to technical progress should itself be detached from same. The call is for the legislator to provide the appropriate legal framework.

Whether and to what extent AI generated inventions will become an issue in future infringement litigation remains to be seen. While computer-implemented inventions are a common object of such, it will take some time for infringement cases concerning inventions generated by AI to be tried before the courts. For now, the most significant hurdle remains to validly file the patent. That being said, the issue of inventorship would not be the first seemingly formal issue to evolve into a much-debated additional battleground in an infringement litigation.