Artificial intelligence (AI) has become one of the main topics of debate in the public space in recent years. AI has numerous implications for countless areas, ranging from increasing the efficiency of industrial production to drug development and language translators or other technological, scientific, or social applications. Because of its high strategic value and the broad range of possible applications, AI is also becoming an issue of national importance. Countries and private companies worldwide are actively investing significant resources in developing their own AI software with the promise of new tools for reinforcing and streamlining already existing technologies. For example, by 2030, up to 45% of global revenues are expected to be linked to the use of AI. [1] The importance of AI applications will grow, and actors unable to use AI properly could suffer from losing the strategic advantage that AI offers. One such actor invested in AI development is the European Union (EU), which is doing so through the unification of European policy on AI, shared planning, and various kinds of EU-scale investments. However, even though the EU is actively involved in AI development and propagates its strategic value, it seems to lag behind other global players. So, what is the root cause of this seeming inability to match other players? Furthermore, what could it mean for Europe?

Main Global National Artificial Intelligence HUBs

Today, the leading AI research and development landscapes can be found in the United States, China, and Europe. This state is also matched by the level of investment in AI in those areas. In the last few years, the United States and China account for approximately 75-80% of global AI investment (76.6% – 2022, 76% – 2021, 80% – 2020). [2] If nothing else, this development reflects the growing interest of global players in AI. However, in this context, focusing on these investments‘ original motivations and output is also essential. Information on where the investments are going and what is their intended purpose can also provide a relevant explanation of Europe’s position. According to the 2021 Artificial Intelligence Index, the quality of research in Europe is comparable to that of the United States and China in many ways. Especially if we look closer at indicators like the quantity and quality of various publications. [3] Also, the amount of national funding is comparable. However, the data itself rarely reports on how the funding is used. 

China has long viewed the development of AI as a national priority. It has established a framework of regulations that influence the activities of central and local governments, the private sector, and academia in AI. [4] This framework thus allows the Chinese government to guide the development of AI and, most importantly, to gain greater control over the development and output of research. In other words, financial flows dedicated to AI research can be redirected in China quite easily to work in favour of the strategy, usually determined by the government reports, China’s State Council, or Politburo itself. Such distribution of funding is next to subsidies like state-sponsored competition awards, done through annual Government Guidance Funds (政府引导基金). Governments on central and local levels establish these funds to support the sectors whose development is pursued by the central government. [5] For example, in 2022, a total of 2,107 funding programs in the form of research contract awards with a total value of 12.84 trillion yuan (€1.64 trillion) was announced in the frame of government guidance funds, focused primarily on entrepreneurs and private firms. [6] Still, although the private sector has a long-standing and growing role in the Chinese environment, it remains more of an adjunct than the United States‘ approach. The exception in this case is military procurements, in which the Chinese People’s Liberation Army relies exclusively on the private sector. However, the approach is still driven by the desire to achieve a managed „civil-military fusion“ of AI research in China. [4]

To state that the United States is not paying attention to funding educational and public research institutions would be a bit of an overstatement. At the same time, it is necessary to acknowledge that the funding priority focuses on AI research programs done by the private sector. [3] In particular, a sector significantly funded by the federal government is the defence sector. We can find 95% of the federal government’s AI-labeled funding under the designation NAICS 54 (designation for professional, scientific, and technical services). Although NAICS 54 includes a wide range of disciplines, 84% of the total value of funding under NAICS 54 is linked to contracts related to the Department of Defense (DoD). [7] The federal government’s total investment in AI in 2022 was approximately $3.3 billion (about €3.12 billion). [8] Two main streams of development are observable in the United States. The first stream is scientific innovation in AI, driven by market demand and directed by private companies‘ investments to achieve higher profits. The second stream is the direct funding of projects of strategic value based on decisions by the federal government, which are supposed to complement market-driven investments.

Although China’s and the United States‘ approaches differ in many respects, the two premises are identical. The first involves adjusting research so that its outputs correspond to strategic interests (security or otherwise) while supporting national competitiveness. The second premise is that a significant portion of the investment is directly related to the state’s security apparatus or AI fields, whose application could have direct security implications. These are, for example, decision science or computer vision, which also saw the largest share of funding in the United States in fiscal year 2022. Thus, in both China’s and the United States‘ cases, there is a strong emphasis on AI’s military or civil-military dimension.

Next to the United States and China, the EU represents the third most potent competitor in AI research and application. EU differs from the previous two challengers primarily in its approach to AI. First and foremost, Europe differs in the structure of its investment system. Currently, through the „Horizon Europe“ and „Digital Europe“ projects, the European Commission plans to invest approximately €1 billion per year. [9] We must remember that in the European case, the government investments of the individual member states are also essential in addition to the mentioned €1 billion. For example, Germany has announced a doubling of its investment in AI in 2023, which is expected to reach up to €1 billion per year after the increase. [10] Therefore, up to a similar amount as the investment made by the European Union. Adding up all European multinational and national investments, it is very likely that the nominal amount could be comparable to the amount of government investment made, for example, by the United States. Thus, the problem is not rooted in the nominal amount of finances spent on the research and development of AI but rather in the coordination and management of funding. The United States and China have a framing plan and a clear overview of where the funds are going. Meanwhile, in Europe, direct financing could lose its potential due to problems like the duplication of research or incompatible strategies applied by the member states.

In addition, the European AI environment is especially interesting in the frame of legal regulation and ethics of AI development. Like China and the United States, Europe has its legal framework that defines the objectives and, above all, the principles of research and application. We could even argue that Europe is a world leader in the area of legal principles and research ethics related to AI. It tries to embed the notion of safe artificial intelligence and thus sets a standard similar to what has been done, for example, in data protection or cybernetics standards. [11] However, this area of excellence does not offer a way for Europe to close the technological gap between the United States and China. For example, despite Europe’s proactive efforts in AI, only 786 patents have been registered by the European Patent Office (EPO) for 2022 in the EU. [12] Meanwhile, in the United States, the number of registered patents reached in the same year almost 16,000. [13] In this sort of comparison, there are two weak points. The first is that the quantification of registered patents on the European level by EPO possibly varies from the actual nominal number of patents registered in all EU member states combined. In other words, patents registered by the National Patent Office do not necessarily have to be registered by EPO or vice versa. The second flaw is the exclusion of China from this comparison. Such incompleteness is caused primarily by the impossibility of verifying the number of AI-related patents from China because data are simply missing. Nevertheless, these data make the performance and potential of the European AI environment partly questionable, if nothing else.

Obviously, conditions and approaches to AI national environments vary from country to country. However, these differences are not necessarily based on distinct funding rates or a disregard for AI per se. Also, we must acknowledge that more than one main factor is responsible for the current situation. There are several, and we will examine the most important ones in the following sections.

History Against Europe

Technological innovation has been problematic for Europe in the long term. The origins of this problem cannot be traced to one particular place in history. However, the dynamic technological development in the 1970s and 1980s was crucial for the current AI development situation. [14] This era gave rise to several companies, like IBM or Microsoft, one of today’s fundamental elements of the United States AI environment, followed by technology giants such as Apple or Atari, which, among other things, have set the standards for the computer and computing industry. This complex process prepared the sturdy foundation for the rise of big tech companies at the beginning of the new millennia, such as NVIDIA, Google, Palantir, or Amazon, which significantly contribute directly or indirectly to today’s AI development in the United States. [15]

The European computing industry has undergone a different and less dynamic development process than the one in the United States. While comparing Europe as an international actor, it is essential to comprehend that Europe as a political structure and actor as we know it today only existed in its modern form from the end of the 20th century. Moreover, even at that given time, it was incomplete. Both in shared strategy and missing numerous members, which are part of today’s European Union. Just like in the United States, the course of history influenced the development of the European computing industry.

The eastern part of „today’s EU“ was, until the end of the Cold War, chained in political reality, practically disallowing effective development of technological and scientific know-how if it was not expressed as a matter of national interest. Eastern Europe was not short of scientific personnel or know-how but rather adequate and stable financing, which could allow progressive development in the field. On the contrary, the development in the United States was supported by both the state and, most importantly, the market economy. On top of that, many Eastern European countries had a problem establishing larger technological companies capable of initial research into new technologies, even in the post-Cold War era. A clear example was the state of the software solutions market segment at the time. Very often, small-sized and lowly standardized individual software projects dominated the market. Such solutions were sufficient and cheap; therefore, the market was not lucrative enough for any possible investments, possibly leading to innovation and leaving those countries unprepared for the new incoming technological era. [16] [17]

In Western Europe, the functional features of market and societal conditions influencing computing industry development were closer to conditions in the United States. With this logic, we could expect an industry development curve that, to some extent, correlates with the situation in the United States. However, the reality was slightly different. The key to understanding such reality is simply accepting that Europe, moreover, the EU, is not the same actor as the United States. Even though the idea of a unified Europe was present for a long time, it came into existence relatively recently, and even today, the unity is nowhere close to a federalized United States. This „division of Europe“ into comparably smaller countries influences the state of the European computing industry today. Before the European Union existed, there was no „convincing evidence that Western Europe is relatively backward in converting technology into economically efficient innovations“. [18] This condition persists, but the European national division holds Europe’s global potential back. While we can find numerous large AI industrial and research hubs in the United States and China, the European AI ecosystem consists mainly of small or medium-sized companies or startups. Despite the efforts of the European community, the market is still partly fragmented, and the approach to AI development is influenced by national biases toward regulation, legislation, and taxation, just like in the case of any other technology. [19] The ability of Western Europe (and later the European Union, involving Eastern European countries as well) to be the strong competitor is curbed by the evolution of European politics and historical legacy, which is partly still present today.

The emergence of smaller European companies, like Aleph Alpha or Mistral AI, indicates a positive course of development. However, because of the scale, these companies can hardly reach the same capabilities as the American industrial complex. Also, the Chinese AI industrial complex, with companies like Baidu, Alibaba, Tencent, and Huawei, represents a tough competitor for Europe, which is missing companies of such scope. [20] We cannot expect the European AI environment to be able to develop in just a few years. For example, the roots of the current Chinese AI environment are in the 1970s, coming together with reforms in the approach towards scientific research. On top of that, development in China from the beginning of the 21st century has also been supported by both rapid economic growth and the influx of new knowledge coming into the country. [21] Even though we cannot exclude the possibility of industrial espionage, a large amount of knowledge came into China thanks to China’s official activity outside of Chinese borders. In this stance, we talk, for example, about outreaching organizations attracting foreign scientists, recruiting foreign „talents“, activity at scientific exchange forums, investments in AI research abroad, and many more. [20] Gradually, over the past few years, AI has become a strategic priority for China while getting to a level at least somewhat comparable to the United States.

Europe has never experienced a „European way“ of the evolution of the AI environment like the United States or China did. However, the phenomenon observable in recent years could indicate a change. The number of newly founded AI startups in the EU annually reaches a similar quantity to that of startups founded yearly in the United States. [22] This is undoubtedly a positive sign indicating the potential for the EU to become a global player, possibly leading to the creation of a solid and resilient AI industry of European type. However, the less developed computing industry ecosystem and, most importantly, the still present scattered political nature in Europe slow down this process. To match the current state of the AI industry in the United States and China, Europe needs to support development at an EU-wide level since fragmentation in such a crucial sector as AI has proved inefficient.

Symbiosis of Motivation and Financing

As previously stated, funding management is as important as the extent of funding. Between 2007 and 2020, 73% of all funding in the EU went to education and research institutions. [23] The majority of European research on AI is done in universities, which could be problematic in the long term because of the probable lack of monetization potential. In comparison, the private sector is motivated by market demand and, therefore, has a higher chance of return on investment. Usually, the problem is the initial investment, which could help develop the European AI landscape much faster and could be covered by EU financing. Although the approach in Europe is starting to change in the name of competitiveness, and funding is increasingly leaning towards market-focused innovation, Europe is again a few steps behind. In 2017, China released its new AI strategy, the „New Generation Artificial Intelligence Development Plan“. One of the principles of this plan is the industry’s primary focus on the market. [24] In the United States, the strategic importance of AI has been articulated even earlier, since 2016. [25] Moreover, in 2018, AI has been identified as one of the United States national security priorities. [26]

In Europe, the market value of AI technologies has only started to become a priority in the last 2-3 years or so. This change has been driven mainly by the articulation of the strategic importance of the sector by some European national actors. This change in thinking about the strategic value of AI resulted in the 2018 „Coordinated Plan on AI“, which was updated and brought into action in 2021. [27] The shift in the EU’s approach, also called „A European approach to excellence in AI“, proclaimed the importance of „making the EU the place where AI thrives from the lab to the market“. [9] As mentioned earlier, this change in thinking has resulted in an investment of €1 billion annually. However, the amount of investment is questionable in this respect. It may limit not only the actual research output but also the speed at which the research and production capacities of the EU respond to trends in other parts of the world. [28] It can be countered by the fact already mentioned above that the combined level of European investment may not be so markedly different from that of China and the United States. However, the question is the symbiosis of the motivations of individual European countries among themselves and in relation to the motivation of the EU. Although the funding system has flaws, it is fair to say that Europe is not necessarily in a position of inferior quality or performance but in a time crunch and administrative chaos caused by the European fragmentation of opinions, which makes it challenging to catch up with the lead of the United States and China’s high-performance AI environments.

In the context of government investment, the motivation of the United States‘ economic and military-strategic rivalry with China may also partially explain Europe lagging behind. It has already been mentioned that a significant part of the investment in both China and the United States is directly or indirectly linked to military research and procurement. At the same time, the market itself heavily supports the private sector. The aspiration to become or maintain a position of global hegemon can be the appropriate motivation to increase investment in a strategically important industry such as AI. This situation may naturally have an accelerating effect on the scale of investment to support the development of a national AI environment. In this case, Europe lacks its peer-to-peer rival, who would act as a motivator. Suppose we focus on the civil-military fusion of AI research in Europe. In such a case, we could conclude that „there is currently too little European thinking about what artificial intelligence means for the military“ and security. [29] However, this trend seems to be gradually changing. The exception in Europe lagging in approach to militarized AI is France, which considers the military dimension of AI and the need for symbiosis and cooperation between civilian and military development and application of AI systems. [30] However, the perception of the issue from a French perspective, in contrast to German funding, whose primary motivations are economic, again underlines the issue of funding management at the European level.

The motivation and related financial management create a challenging obstacle for future European AI development. Europe needs a shared vision and motivation regarding AI development. The approach at the EU level thus gives the impression that Europe, as a coherent actor, seems to react to AI just to respond to AI phenomena rather than genuinely knowing how and what it wants to achieve with it. Several European countries offer favourable conditions for the development of AI. However, without future political and economic mobilization at the European level, neither Europe nor its members can compete globally with actors such as the United States and China.

EU the Regulatory Hegemon

Another potentially influential factor that may have a particular impact on the future evolution of the gap between the EU and the other two leaders in the AI race is the European regulatory environment. The EU has long been moving towards a regulatory approach that is accountable to citizens and consumers. However, such an approach is also negatively reflected in the research and deployment of AI systems. In this respect, introducing the EU AI Act is particularly important and raises some doubts. The proposal defines that all European AI should be „safe, transparent, traceable, non-discriminatory and environmentally friendly“. The debate is to what extent these essentials, together with the regulation of the riskiness of AI systems, affect the difficulty and cost of research and application. An essential part of the regulation is also the provision saying that all AI systems should always be under the direct control of individuals. [31] In this manner, regulation essentially rejects AI systems‘ general autonomy, affecting how AI is used and developed.

The question of regulation is, of course, not only present in Europe but also in certain forms in the United States and China. However, each actor approaches regulation differently and with different desired outcomes. United States‘ legally binding regulation environment is currently practised primarily through presidential executive orders. The latest one from the end of 2023 sets certain principles regarding the use and development of AI. This order is based on earlier executive orders and the legally non-binding „Blueprint for an AI Bill of Rights“ made by the White House Office of Science and Technology Policy. The latest executive order expects AI to be safe and secure while ensuring responsible innovation, competition, and collaboration, as well as supporting domestic workers and attracting foreign experts. It also highlights the need to protect the privacy, civil liberties, and interests of the American population based on existing legal protection against discrimination, fraud, and other harms that could arise from the use of AI. One of the interesting final principles is emphasizing pioneering in global societal, economic, and technological progress in AI development. The United States regulations follow similar looser rules regarding protecting individuals and the legal system. However, at the same time, it puts extensive pressure on improving the American AI complex to be globally competitive. [32] [33] Besides, many large companies have their own AI ethics code; nevertheless, such codes are not somehow binding.

On the other hand, Chinese regulations are more specific than those of European or American ones. Regulation in China is prepared by the Cyberspace Administration of China (CAC) and composed of numerous regulatory acts focusing on specific AI topics like algorithm recommendation, generative AI, or ethics of development. Besides „traditional“ regulations regarding protecting privacy and developing safe AI, regulations focus primarily on state control over AI. It means, for example, providing information about used algorithms, verifying any applications using AI before releasing it to online stores by a national authority, verifying a user’s real identity before using any generative AI application, and so on. [34] The whole process of regulation in China mainly focuses on a firmer grip over AI and its use for national, often ideologically underlined purposes. In the case of the currently most developed branch of AI, the generative AI, CAC even ordered any generated content by this AI needs to be in line with China’s „core socialist values“ while using legitimate data sources, which confirms AI’s additional ideological role for China. [35]

The regulation content differs based on the role of AI for the regulator. The general scope of regulation in Europe considers primarily the protection of the consumer and society itself. However, if done incorrectly, regulation could severely affect Europe’s competitiveness, which is not considered overly crucial in current regulatory codification, at least compared to the content of United States executive acts. The European bureaucratic burden to meet regulatory requirements could slow the application of AI systems. An even more critical aspect is how the content of regulation could affect the development of AI systems. Unless the content of the regulation changes, numerous projects, especially in the field of generative AI, could see an increase in its complexity and, therefore, an increase in the financial and time requirements for the development of these systems. [36] As a result, this could limit the ability to produce certain types of AI systems in Europe. At the same time, companies outside of Europe that are unaffected by European legislation could potentially supply such products to the global market at the expense of European companies.