Analyzing Global AI Dependencies through Intellectual Property Rights

Policy Brief

Executive Summary

Artificial intelligence (AI) has become an important strategic asset in foreign policy. As a general-purpose technology, AI can increase a state’s economic power and enhance its political influence. It has also become an element of protecting states’ national security and defense interests. Therefore, access to AI technologies and the ability to participate in AI innovation is key to increasing the economic competitiveness and sovereignty of states. Where companies gain importance within a country or region by virtue of their innovative capacity, technological dependencies can arise. At first, such dependencies on foreign AI technologies become visible in the private sector, but eventually, they can develop into political ones. Ultimately, this can lead to shifts in the global balance of power and result in geopolitical tensions.

To uncover strategic dependencies on foreign companies, countries, or regions in the field of AI, European policy makers need to examine whether European industries are highly dependent on foreign AI technologies. Only if the European Union (EU) knows its position in the global AI innovation ecosystem, it can develop sound technology foreign policy. It needs to identify leading AI innovators, but also reflect on its dependencies and weaknesses that prevent it from achieving a higher degree of autonomy, especially when it comes to national security or digital transformation.

This paper provides European foreign policy makers with in-depth information on how to uncover the EU’s strategic dependencies in AI. It applies a particularly revealing technique of identifying dependencies in AI: It maps the AI ecosystem with respect to innovative capacity. This can be measured by examining different intellectual property (IP) rights that allow for the protection of inventions in AI. The distribution of AI patents, for example, is not only a strong indicator for the innovative capacity of companies but also more generally for the inventiveness of states and regions. Currently, companies from the US and Asia dominate European patent applications in the field of computer and digital communication technologies, to which AI technologies are of fundamental importance.

However, it is not enough to count patent applications. Rather, as a first step in mapping Europe’s AI landscape, this paper suggests examining the distribution of IP rights along the core industrial inputs of AI. This allows an understanding of the innovative capacity of AI companies and the general structure of AI ecosystems; it also helps to identify locational (dis)advantages that arise from the lack of international harmonization of IP rights protection.

In addition to patents, tradesecrets and copyrights are important indicators for measuring innovation in AI. Moreover, European foreign policy makers need to understand what AI is and which individual elements make up this technology. This requires a breakdown of AI into its key industrial inputs: algorithms, hardware, and data. These categories—patents, trade secrets, and copyrights, as well as algorithms, hardware, and data — form the analytical matrix this paper applies to identify the ways in which innovation in AI can be protected by IP rights and how this, in turn, can contribute to technological dependencies.

This paper has identified three main challenges related to the IP rights protection of AI’s core industrial inputs.

  • U.S. patent law offers more room than European provisions in terms of opportunities to patent AI algorithms. This could represent a locational disadvantage for the EU, especially considering that algorithms form the core of AI innovation.
  • The project of mapping Europe’s AI ecosystem is further complicated by pending court cases. In the EU as well as in the US, there is legal uncertainty as to whether private sector companies may train their AI models on copyrighted data. The situation might tilt towards AI developers and weaken the legal position of copy-right owners. If at all, this seems more likely to happen in the US, if US courts should rule analogous to the Google Books decision, which allowed the company to scan millions of books that had previously been under copyright protection.
  • In the case of AI hardware, the challenge does not lie in the different application of IP rights, but rather in the lack of a European hardware infrastructure in the form of hyperscale cloud platforms or supercomputing clusters. Therefore, European AI companies and increasingly European cutting-edge research are already dependent on foreign, especially U.S. cloud computing infrastructures.

In sum, examining different IP regimes along AI’s core industrial inputs highlights the need for the EU to engage in a systematic mapping of this complex space to identify its strengths, weaknesses, and strategic dependencies. This constitutes a first step towards building long-term strategic capacities in European AI technology.


The SNV’s Artificial Intelligence and Foreign Policy project was made possible by the generous support of the German Federal Foreign Office and the Mercator Foundation. The views expressed in this paper do not necessarily represent the official positions of the German Federal Foreign Office or the Mercator Foundation.

Erschienen bei: 
Stiftung Neue Verantwortung
05. Mai 2022

Philippe Lorenz