EXECUTIVE SUMMARY
The Artificial Intelligence Act (Regulation (EU) 2024/1689 or the “AI Act”) integrates into the European legislative framework with the aim of establishing a harmonized set of rules for the use of artificial intelligence within the European Union. Proposed in 2021 by the European Commission and adopted on May 21, 2024, by the Council of the European Union, the AI Act is a fundamental part of the EU’s digital strategy, aiming to promote innovation and ensure safe and responsible use of AI technologies. It provides, indeed, the EU with the most advanced regulation on Artificial Intelligence (“AI”) among all the major geopolitical players in the world, aiming at ensuring the generalized application to all AI systems throughout the main industries and the development of safe and trustworthy AI systems.
While some countries have set forth rules to deal with specific issues where AI is involved, the AI Act takes an innovative - and more wholesome - approach. It provides a specific definition of AI, which delimits its scope of application. Then, it classifies specific developments of AI systems into four levels of risk (i.e. the risk-based approach), each one imposing specific duty to developers. AI systems classified in the unacceptable risk category (such as governmental social scoring) are prohibited. High risk systems must be closely monitored by the European institutions and must be subject to risk assessment and risk mitigation practices. Limited risk systems must undergo public disclosure of relevant information to enhance their transparency. Lastly, minimal risk AIs (the category that counts most types of AI) remain free to use. The international nature of the AI Act and its all-embracing regulatory ambition aim at influencing positively other international players such as the United States and China.
The AI Act, being fully and directly applicable in all the MSs, interacts with other regulations and provisions such as the General Data Protection Regulation (“GDPR”), the Cybersecurity Act, and the Digital Services Act, creating a coherent and comprehensive regulatory environment covering various aspects of digital technologies and data protection. In addition, few States already have some kind of legislations that directly involve AI, Greece being the most remarkable example. Others have a mix of different legal sources that directly or indirectly influence the development of AI, setting forth provisions that might impact the AI future development. Moreover, some States have already taken action to complement the AI Act’s provisions, introducing a stricter regime for some specific technologies while others like France and Spain, are still considering the need of introduce complementary legislation to avoid risks related to errors in AI-generated decisions, or cybersecurity issues.
While individual countries within the European Union were not directly providing funding specifically earmarked for compliance with the AI Act itself, as it primarily focuses on regulatory frameworks rather than direct funding initiatives, countries like Switzerland, Sweden, France, Germany and Hungary and the EU as a whole have been actively investing in AI-related research, development, and innovation through various programs and funding mechanisms. These initiatives are generally aimed at supporting the broader digital transformation goals of the EU, which includes enhancing AI capabilities and competitiveness.
But what is the future of the concrete application of the AI Act? To answer this question mark, it is needed to dig into the preliminary characteristics of individual States as well as their potential intervention in terms of finance by asking: (i) Do States already regulate AI at national level? (ii) Do they plan to complement, or even enhance, the AI Act’s provisions? (iii) Do they want overall to financially support the development of AI systems?
The following analysis provides a critical overview of the most relevant countries in Europe, namely Albania, Belgium, Bulgaria, Cyprus, France, Germany, Greece, Hungary, Italy, North Macedonia, Portugal, Romania, Serbia, Sweden, Switzerland and Spain.
***
Albania
Albania is gradually integrating AI into its digital programs to enhance efficiency and innovation in the public sector. Law 87/2923 ratifies the agreement for the adhesion to the “Digital Europe” program, which focuses on advanced digital skills in general, though specific details on AI are not set forth. Law 43/2023 regulates electronic public services that do not require the physical presence of the applicant, while the Decision of the Council of Ministers No. 370/2022 approves the Intersectoral Strategy “Digital Agenda for Albania”, which promotes AI to improve public services, AI integration in the public sector, and data management. It also established a High-Performance Computing Center. Unfortunately, official information on governmental funding of AI is not yet available.
Belgium
In Belgium there isn’t a specific regulation on AI. However, Collective Bargaining Agreement No. 39 is applicable to these systems. The bargaining requires businesses with more than 50 employees to preemptively inform and consult trade unions, if they invest in technologies that impact on 10% of the personnel, or more. Additionally, the collective labor agreement enacted by the Royal Decree of 21 February 2024 in the banking sector makes it mandatory to train workers in the use of AI. Additionally, acts and resolutions adopted by the Belgian legislative branches regulate the use of AI in sectors such as defense, public administration and automotive (especially relevant is the permission to run tests on automated vehicles). Government funding is split into regional and state levels. The Flemish region allocated 32 million euro for AI Research and Development. It also set forth fiscal incentives for supporting the cost of labor. On the other side, the Walloon region launched the “Application and Research for Trusted Artificial Intelligence” project for the development of AI systems that enhance the competitiveness of Walloon’s businesses and awarded it 32.3 million euro. Brussels also offers many grants and subsidies to businesses in the AI industry. These three regional levels all complement the Federal programs, which mostly invest in defense.
Bulgaria
The Bulgarian Government adopted the “Concept for the Development of Artificial Intelligence in Bulgaria until 2030”, which recommends conducting a detailed analysis for a legislative proposal to be carried out only after the entry into force of the AI Act. No proposals will be forwarded until then. In the meantime, Bulgaria plans to use the European Structural Funds, the “Horizon 2020” programs, the National Recovery and Sustainability Plan, and the state budget to fund projects in the AI sector. Moreover, Bulgaria established the “Institute for Computer, Science, Artificial Intelligence and Technology” (“INSAIT”) and invested 85 million euro in it through the Ministry of Education and Science. The INSAIT is set to be the most advanced research institution in Eastern Europe, as it can rely on world-class partnerships with the most important players in the field, such as Switzerland’s “ETH” in Zurich and “EPFL” in Lausanne (two prestigious technical universities), as well as donations by Google, Amazon Web Services, DeepMind, SiteGround, VMware and other tech entrepreneurs.
Cyprus
Cyprus declared that the AI Act will be the first legislation that directly regulates AI. It presented a Strategic Plan for 2024-2026 which has allocated 282 million euro until 2026 for digitalization in general, including AI. However, much information on the development of AI in the country is lacking.
France
France’s own regulation of AI systems mostly revolves around the GDPR, and the “Loi Informatique et Libertés” of 6 January 1978, which apply to the processing of personal data. However, France is one of those countries that also legislate on more specific applications of AI systems, namely automated decisions. Individuals subject to those decisions have a right to obtain a human review of their personal situation, to express their point of view, to obtain a precise motivation in support for the decision, but also to challenge it (in line with Article 22 of the GDPR). On 12 September 2023, French representatives of the Assemblée Nationale proposed a legislation to regulate copyright-related issues, but the proposal was deemed unfeasible, and it was blocked. In 2021, President Emmanuel Macron launched its France 2030 plan, investing 2,5 billion euro in AI Research and Development. On 19 September 2023, the Government established the “Generative Artificial Intelligence Committee” to provide better guidance to the Cabinet for its decisions in the AI sector.
Greece
In Greece, Law 4961/2022 covers multiple technological developments, including smart contracts, the Internet of Things and AI. Only a few modifications are required for the Law to be fully compatible with the AI Act, which testifies Greece’s ability to legislate carefully in the area. In terms of Government funds, the Ministry of Digital Governance issued between 150 and 200 million euro for the development of AI.
Germany
Germany declared that the AI Act will be the first legislation that directly regulates AI. In 2018, Germany adopted a federal strategy for AI, significantly increasing the invested funds in the sector. Indeed, the Federal Ministry of Education and Research annual budget for technology investment increased by 20 per cent since 2017, reaching 1,6 million euro in the current legislature. These funds will be invested in eleven specific areas, including enhancing research to be a driver of innovation, setting up an agenda to expand the AI infrastructure, building the AI infrastructure, promoting social dialogue with stakeholders to responsibly integrate AI into Germany’s institutional systems, and ultimately drafting and enforcing a more effective, innovative and agile AI regulation, that fosters innovation.
Hungary
In Hungary, Government Decree 1573/2020 sets forth the national strategy on AI. This strategy has been developed by the Artificial Intelligence Coalition, which comprises more than 70 Hungarian and International companies, universities, and research groups. The Coalition’s main goal is to place Hungary as a leading power in AI, and to strengthen national businesses’ competitiveness with it. The strategy focuses on transformative programs in the manufacturing, healthcare, energy and logistics sectors. It aims at enhancing Hungary’s technologies and infrastructures, while also fostering the public comprehension of AI and its applications. The Hungarian government, through the venture capital program “Széchenyi” made funds available to support the development of AI. The funded projects include the “AI Innovation and Competence Centre and Data Asser Management”, which promotes the application of AI systems to Hungarian small and medium enterprises, the development of AI-based technology for various purposes and the establishment of the national “Laboratory for Artificial Intelligence”.
Italy
In Italy, the GDPR and copyright legislation are the main tools used to regulate the application of AI and its functioning. In particular, these are significantly important as far as the training process of AI models is concerned, since training sets may incorporate and process personal data or copyrighted work, triggering legal obligations, prohibitions and sanctions. Additionally, Italy plans to further regulate the topic, complementing the upcoming AI Act. Indeed, the Italian Government recently proposed to the Italian Senate the introduction of a new law (Disegno di legge 1146/2024 - “DDL”), which is currently being discussed by the relevant committees. The announced goal is to strike the right balance between the protection of human rights on the one hand, and the development of innovative and beneficial tools on the other. In other words, the DDL dictates general principles to follow, in order to prevent AIs from negatively impacting on human autonomy, excluding humans from the processes in which AI is employed, and depriving them of their individual decision power. Thus, the DDL safeguards freedom of information and communication, fairness in the use of data, and transparency in the training of models. The DDL also proposes to introduce more sector-specific provisions, authorizing, for example, the use of AI systems in public administration and in intellectual professions only in secondary activities in support of the core one. The Italian Government has allocated 1 billion euro to the acquisition of shares in Italian innovative businesses, including those that develop AI systems. It also plans to facilitate partnerships between the public and the private sector, for the development and the employment of fair and efficient AI systems in the public administration.
North Macedonia
In North Macedonia the main regulation that deals with AI is the national Data Protection Law, which is harmonized with the GDPR, while no other specific proposals targeting AI are planned. In 2023, the Government launched “ADA”, an AI-powered digital assistant, which provides information to citizens and businesses. Moreover, the Governmental “Fund for Innovation and Technology Development” has invested approximately 138 million euro since 2013, to support start-ups and innovative projects in North Macedonia.
Portugal
Portugal plans to align its legislation on AI with the AI Act. Portugal is one of the first countries to have developed a national strategy. In fact, in 2019 it adopted the “National Strategy on AI”, which fosters innovation in healthcare, agriculture and education. In line with the AI Act, the National Strategy too is based on a risk-based approach that aims at developing ethical and human-rights-oriented AI. Importantly, Portugal’s strategy includes the establishment of sandboxes, which are secure environments where AI systems can be tested, under the supervision of public authorities, before getting released to the public, mitigating risks related to privacy and security.
Romania
The Romanian Senate has started public discussions on 19 March 2024 for a new legislative proposal on AI. As of today, the proposal would prohibit the automatization of the flow of human resources, as well as the use of biometric data of natural persons (except for crime prevention and detection purposes). It is worth highlighting that the Romanian approach might conflict with the AI Act, therefore it might me heavily modified to align with it. In February 2024, the Ministry of Investments and European Projects published Order No. 464/2024 for the approval of a state aids scheme, and a scheme of contributions to support entrepreneurs in the development of advanced digital technologies. These schemes allocate 3 million euro in grants to foster the development of advanced technologies, including AI.
Serbia
As of today, Serbia set forth voluntary guidelines and recommendations in the AI sector, such as the “Strategy of Development of Artificial Intelligence in the Republic of Serbia for 2020-2025”, and the “Conclusion on Adoption of Ethical Guidelines for Development, Application and Use of Reliable and Responsible Artificial Intelligence”. Besides providing guidance for the development of safer AI models, these acts also envisage the enactment of new regulations and the amendment of existing laws to better face the risks posed by AI (such as a new Data Protection Law). Serbia is currently developing its 2024-2030 national strategy on AI, which is expected to be adopted by the end of the year. As Serbia is also preparing to become an EU Member, it is also likely that it will transpose and issue a law on AI that parallels the AI Act. The Serbian Government also financed AI-related projects with 1 billion euro per year, in the 2020-2025 period. The Fund for Science currently funds 12 projects under the “Program for Development of Projects in the AI Field”.
SwedenThe Swedish Government is currently collecting information to draft a national law on AI, that is in line with the AI Act. But Sweden’s most important tool to get ahead in the AI race comes in the form of investments in the AI field, and the focus on the training of human capital. In fact, Microsoft announced a 3 billion euro investment in AI-related projects, and the training of 250,000 people, highlighting the attractiveness of Sweden in the industry. In 2023 the Government established a commission on AI to enhance the competitiveness of businesses in the sector, which concluded that the public administration would benefit approximately of 14 billion euro per year thanks to the integration of AI in its processes. Sweden also established “AI Sweden”, a government-funded national center for the application of AI, in partnership with entities in both the public and the private sector, as well as universities.
Switzerland
As of today, Switzerland chose a sector-specific approach to AI regulation, including the federal data protection law of 2020 which disciplines automated decisions, granting enhanced transparency and other individual rights that are similar to those granted under the GDPR. However, the legislation might soon be integrated by a new tool, as the Federal Council started an analysis to develop general legislation on AI, by 2025 (which is set to consider the norms of the EU and of the European Council). The goal is to provide Switzerland with a clear and AI-compatible normative landscape. Additionally, the Swiss Federation supports digitalization and technological development projects, including AI programs, in partnership with universities and research institutions, such as the “ETH” in Zurich, or the “EPFL” in Lausanne. Also, “Digital Administration Switzerland” and “AI Switzerland” promote innovation and training in AI-related fields.
Spain
Spain’s main tool to regulate AI is the “National Strategy for Artificial Intelligence 2024”, in line with the AI Act. The Spanish strategy mainly focuses on cybersecurity, which is an essential element for the digital transformation of the economy. In fact, Spanish institutions are currently discussing a Cybersecurity Law, which is set to enter into force in late 2024, to grant a better protection of digital systems. Additionally, Spain established the “Spanish Agency for Artificial Intelligence Supervision” (“AESIA”), with the purpose of granting ethical and transparent practices in the use of AI. The National Strategy allocates 1,5 billion euro – from the National Recovery and Sustainability Plan – and 600 million euro from the State budget, which have already been mobilized. The plan includes investments in supercomputing, the development of language models in Spanish and other co-official languages, the promotion of talents in the AI field, and the expansion of AI in the public and private sectors, with special attention for small and medium enterprises.
United Kingdom
The United Kingdom heavily relies on the GDPR to regulate AI (which also influences automated decisions). The “Artificial Intelligence Regulation Bill” was a tentative to provide the country with horizontal regulation on AI, although its progress was stopped due to the UK’s early elections. However, the Government has already launched several initiatives in support for AI. The “AI Sector Deal” of 2018 invested 1 billion pounds for Research and Development, training and ethical research on AI. Additionally, the “AI and Data Grand Challenge” provides businesses with 2,6 billion pounds from the public sector, and 3 billion from the private one. The “United Kingdom Research and Innovation” program further invested 300 million pounds for proper computing, 250million pounds for the development of AI in healthcare and zero-emission economy, and 100 million pounds in academic centers for technological talents.