Artificial Intelligence Systems as Inventors?

The question of whether artificial intelligence (AI) can be deemed an inventor under patent law has been litigated world-wide. Previously, the US District Court for the Eastern District of Virginia, the Intellectual Property and Commercial Court of Taiwan and, most recently, the UK Court of Appeal dismissed the claim that a non-human entity can be recognised as an inventor under the respective national patent law. An analogous question is currently pending appeal at the European Patent Office. In view of this emerging case-law, the Max Planck Institute for Innovation and Competition (Munich, Germany) issued a Position Statement commenting on the recent decision by the Federal Court of Australia (FCA) that acknowledged an AI system DABUS as the inventor under Australian patent law.

In response to the arguments provided by the judge of the FCA, the authors of the Position Statement argue that the literal interpretation of the term ‘inventor’ as an ‘agent noun’ is insufficient to conclude that non-human entities can be deemed inventors for the purposes of patent law, that the silence of patent law regarding non-human inventors cannot be interpreted as implicit recognition, and that denying the inventor entitlement for AI systems would equally contribute to legal certainty. The authors of the Position Statement highlight that, while the decision of the FCA appears ground-breaking at first sight, it was mostly based on the unverified assumptions regarding the technical capabilities of AI systems in general and DABUS in particular. The prima facie recognition of DABUS as an ‘autonomous inventor’ can promote public confusion about the factual basis of the court decision and, more broadly, the state of art in AI research. In addition, it can instigate public resentment against the unjustifiably granted patent rights, given that patents impose welfare costs in the form of prices above a competitive level.

The legal consequences of the judgment are potentially far-reaching. The FTC held that, pursuant to Australian law, the one who owns and controls DABUS and owns the computer on which DABUS operates can also claim ownership in an invention generated by using DABUS as a tool. However, the FCA decision does not draw a distinction between the assumed capacity of AI systems to invent autonomously and the application of AI systems as computational techniques based on mathematical optimisation and statistical principles in the process of developing an invention. In this regard, the judge’s assumption that the recognition of AI systems as inventors would promote innovation appears speculative and highly dubious. Rather than promoting innovation, the decision might have a discouraging effect on third-party developers of case-specific AI applications and AI-based products and services.

Overall, it is argued that, if at some point AI systems might be able to invent autonomously, the question that should be addressed in the first place is whether, from a social welfare perspective, inventions generated without human input should merit patent protection at all. While the Position Statement responds primarily to the decision of the FCA, the arguments put forward can be of relevance for any jurisdiction dealing with the question of whether an AI system can or should be deemed an inventor under patent law.

Josef Drexl is a Professor at the Max Planck Institute.

Reto Hilty is a Professor at the Max Planck Institute.

Daria Kim is a Researcher at the Max Planck Institute.

Peter R. Slowinski is a Researcher at the Max Planck Institute.