What is the role of science in the EU AI policy approach?

Science throughout the policy journey

The exploration phase, roughly from 2018 to 2020, laid the groundwork for EU AI policies. Years before AI dominated the headlines, JRC scientists were focusing on understanding the implications of AI technologies on society, as well as its opportunities and development trajectories. In these early stages, it was crucial to provide the vocabulary for achieving a common definition of AI, its risks, and the guiding principles. The 2018 Coordinated Plan on AI set up a knowledge service, co-led by JRC, to monitor the development, uptake and impact of AI across Europe: AI Watch. JRC scientific and technical contributions, such as in-depth comparative analyses of national AI strategies, informed major European policies and national strategies on AI of several Member States.  

With the White Paper on AI, presented by the European Commission in 2020, excellence and trust became the baseline of the European approach to AI. In the transition phase, covering 2021-2023, the policy focus shifted to translating principles into regulatory requirements. JRC supported policymakers in defining key concepts such as the AI system, general-purpose AI, foundation models and generative AI. Science also clarified transparency criteria, cybersecurity requirements, and compliance standards for high-risk AI systems. This work led JRC scientists to step in with their expertise during some of the most critical stages of the AI Act negotiations among the European Commission, the European Parliament and the Council of the EU. While AI use went from niche to mainstream with the arrival of generative AI  available to the wider public, the scientific support ensured AI policies to be technically sound and future-proof.   

The adoption and entry into force of the AI Act, and the launch of AI Continent Action Plan, Apply AI and the AI in Science Strategies in 2024-2025 represents another stage of EU AI policymaking, the maturation. In this phase, science is building the evidence that makes compliance achievable and adoption possible by individuals and organisations.  

One example is the meaningful uptake of AI in the public sector. JRC scientists propose an AI adoption framework for EU public administrations under the Apply AI Strategy. The framework offers practical guidance for civil servants and stimulates the debate with Member States on the effective use of AI by the public sector.   

Another example is JRC research on the technical standards needed to make the AI Act's requirements work in practice. This work helps providers and deployers of high-risk AI systems to understand and navigate the requirements, reducing efforts and uncertainty especially for small and medium-sized enterprises.