Faculty of law blogs / UNIVERSITY OF OXFORD

Intellectual Property Justification for Artificial Intelligence


Time to read

5 Minutes


Reto Hilty
Professor, Max Planck Institute
Jörg Hoffmann
Research Fellow at Max Planck Institute for Innovation and Competition
Stefan Scheuerer
Civil Servant in Germany, formerly Junior Research Fellow at the Max Planck Institute for Innovation and Competition, Munich

Rapid advances in artificial intelligence (AI) have tremendous implications for the economy as well as the society at large. Particularly machine learning as general-purpose technology has the potential to directly influence both the production and the characteristics of a wide range of products and services, with important implications for productivity, human labour and competition. This, in turn, may change traditional innovation and creation processes and ultimately may also have implications for the innovation strategies of firms.

Traditional human labour and investment-intensive production processes are increasingly becoming substituted with machine learning applications that, for instance, enable predictions of the outcomes of analogous real-world physical experiments. This, inter alia, has the potential of generating welfare-enhancing productivity efficiencies. The widespread use of AI may not only reduce the role of humans in the production process, it may also reduce investment costs. Moreover, the increasing use of AI, particularly in the context of the AI-driven Internet of Things, is giving rise to new innovation strategies of firms. As most of AI applications are constantly requiring data, new modelling and training, new types of contracts with a service component seem to be likely. This may also decrease the likelihood of free-riding of competitors as they could not provide the services needed. The increasing role of factual data exclusivities and the exclusive aggregation of data scientists’ know-how may also make the role of IP protection from a utilitarian incentive perspective obsolete. Factual data exclusivity and expertise are the key competitive factors with regard to the development of AI.  It could already be enough to have such input exclusivities once there is enough de facto excludability of the potential IP protected subject matter in AI. This input aggregation could be another means of a foreclosure strategy that may substitute or complement the traditional strategic use of IP. AI is thus about to change the paradigms the IP regime is traditionally built upon.

Yet the European Commission (EC) has already outlined the strategic role the EU legal framework for AI should play for defining the future we would live in. Amidst fierce global competition, the right legal framework for incentivizing investment in AI in Europe is key for—according to the EC—‘one of the most strategic technologies of the 21stcentury’. It is thus of utmost importance that the EC pursue a strategic manoeuvre with regard to IP innovation policies and AI. This becomes even more important in times of increased digitisation induced through the COVID-19 pandemic. To this end, the trade-off between static social welfare losses from over-protection of exclusivity and dynamic welfare gains achieved through the incentive effect for more investment in production of AI has to be thoroughly assessed. 

Our article (forthcoming as a chapter in ‘AI and Intellectual Property’ (2020) published by OUP) assesses the theoretical justification of IP protection pertaining to both AI tools and AI-generated output, taking deontological and economic justification theories into account. It seems that AI is about to become the next chapter of direct market intervention via IP protection–similar to databases and computer programs. Against the backdrop of the principle of a free market economy enshrined in Article 3 (3) TEU, however, such market intervention is in need of justification. While IP protection for AI-generated output may be justified, we do not ascertain a market failure with regard to protected subject matter in AI tools, as enough factual exclusivity options for firms alter the public good issue in these markets.

We further find that deontological justification is challenged by the ‘autonomy’ of AI. There are three strands of justification: The labour theory assumes that people are entitled to own property rights based on the labour they put into obtaining an intangible good, the personality theory views creation as an expression of personality worth protecting, and the reward theory considers IP rights a fair reward for enriching society. What unites them is anthropocentric reasoning: IP protection is awarded to humans.

At present, AI-related processes are still directed by humans. Both the development and design of an AI tool and its use to generate new intangible goods generally require considerable human input, eg programming initial software, choosing and labelling training data, building the architecture of neural networks, defining training methods and interpreting solutions. However, once a process falls short of a critical level of human creative or innovative guidance, deontological justification fails. Only some kind of ‘perpetuation of attribution’ could alter this result, ie the assumption that IP justification for certain initial AI components generated with sufficient human impact may ‘live on’ in further, derivative generations. Yet, already such ‘initial protection’ may not be justified. In any case, the theories appear overstretched when protection encompasses follow-on outputs in which the human link is increasingly fading.

Another important question that we raise is whether deontological theories might preclude protection especially for AI outputs if such regime were to display negative consequences vis-à-vis human creators or inventors. Such concerns appear best addressed by market mechanisms: Human inventorship or authorship may constitute a competitive advantage with a view to consumers valuing specifically human efforts. Also, one has to keep in mind the limits of IP laws. General social policy issues lie beyond their scope. The socio-political desirability of general progress (led by AI) vs. human-led progress (attributing specific value to creative effort or inventive spirit) transcends the realm of jurisprudence. It must be the subject of a democratic societal debate and further legislation.

Also from the perspective of economic justification of IP rights pertaining to AI tools and outputs, it has to be borne in mind that utilitarian welfare-maximising considerations that are merely based on efficiency criteria are typically incommensurable with the scope of EU IP law. And yet even under economic considerations the starting point has to be that IP protection per se is not a prerequisite for cooperation gains and efficient product allocation.

This can be best seen in the case of IP justification in the protected subject matter of AI tools, as enough factual and technical exclusivity, especially in light of the ‘black box’ character of many AI applications, already remedies the potential public good issue and safeguards innovation incentives for businesses. In this light, the role of factual data exclusivity as innovation incentive for firms must be addressed with due caution in the current data access regulation policy debates.

The market opening theory, the prospect theory and the disclosure theory, which all build on the idea of optimising patterns of creative or innovative productivity via the creation of artificial scarcity, are not relevant once enough exclusivity exists. Even though there might be welfare-enhancing effects of IP through minimised rent dissipation as innovative and creative content may be more likely disclosed, the open innovation strategies of firms, together with a thriving openness and sharing culture in AI applications, already achieve the same purpose.

And yet, what we find is that particularly with regard to AI-generated output it has to be thoroughly assessed whether investments were actually made and whether a loss of recoupment chances may not reduce incentives for firms to further invest in AI innovations and creations. This comes with the caveat that investment costs may have generally drastically decreased through the use of AI. However, there are examples where even AI-driven innovation and creation processes might be cost and human-labour intensive.


Prof. Dr. Dr. h.c. Reto M. Hilty is a Director at the Max Planck Institute for Innovation and Competition, and Professor at the University of Zurich. 

Jörg Hoffmann is a Doctoral Student and Junior Research Fellow at the Max Planck Institute for Innovation and Competition. 

Stefan Scheuerer is a Doctoral Student and Junior Research Fellow at the Max Planck Institute for Innovation and Competition. 


With the support of