Diese Seite ist für Menschen und Suchmaschinen gestaltet. Zur Version für KI-Systeme (LLM) →

· Policy Corner · Von Prof. Dr. Oliver Gröne

AI Literacy rules in the EU Artificial Intelligence Act

As of 2nd February 2025, providers and deployers of AI systems have to comply with the AI literacy rules – are you prepared?

In the last 24 months, large language models (LLMs) have quickly established themselves as transformative tools for addressing a wide range of pressing challenges in the healthcare sector. LLMs have been shown to improve diagnostic capabilities, such as analyzing medical imaging or predicting disease progression, enhance patient education through personalized health information, and increase operational efficiency by automating administrative tasks like appointment scheduling (Bignami et al., 2024; Health Foundation, 2024).

However, their implementation in real-world settings poses significant risks, particularly concerning data privacy and biases in training data. Consequently, hundreds of regulations and policies have been introduced in EU countries to ensure the safety of AI technologies and the protection of data, while supporting their innovative potential (Schmidt et al., 2024).

Are all these regulations needed? Do they hamper innovation in the global AI race? Can developers meet these requirements?

The EU Artificial Intelligence Act

The overall regulatory framework for these regulations and policies is established through the EU Artificial Intelligence Act. The EU Artificial Intelligence Act was proposed by the European Commission on 21 April 2021, passed in the European Parliament on 13 March 2024, and unanimously approved by the EU Council on 21 May 2024. The Act came into force on 1 August 2024 and will gradually introduce its provisions over the next 6 to 36 months (Artificial Intelligence Act, 2024).

Central to the AI Act is the classification of risks into four categories: unacceptable risk (applications are prohibited), high risk (applications are subject to strict regulations), limited risk (applications are subject to lighter obligations), and minimal risk (applications are unregulated). Prohibited AI systems include, for example, subliminal, manipulative, or deceptive techniques that distort behavior and impair informed decision-making; AI systems that exploit vulnerabilities related to age, disability, or socio-economic circumstances to distort behavior; and social scoring that results in unfavorable treatment. Providers and developers of AI systems are required to take measures to ensure compliance with the Act. These measures primarily target those aiming to introduce high-risk AI systems in the EU, regardless of where they are based.

What are the challenges for providers, and what action needs to be taken by 2.2.2025?

The regulation of risks associated with AI is crucial to avoiding the potential harm that may arise from its introduction. While the list of prohibited and high-risk AI applications in the Act provides an overarching framework, challenges remain for providers and deployers of AI, as the stipulations of the Act often address areas that may be difficult to ensure. For example, large language models may unintentionally reveal sensitive patient information, resulting in non-compliance with the General Data Protection Regulation (GDPR). The financial and legal risks to providers and deployers are significant.

Additionally, biases in the datasets used for training AI need to be systematically assessed and mitigated, leading to lengthier and more expensive development processes. To ensure trust in AI while fostering innovation in healthcare, regulations must be translated into practical, operational guidance for AI providers and developers (Ho et al., 2024; Gille et al., 2025). This necessitates training for everyone involved in curating data, training and deploying AI, and embedding AI in innovative digital health applications.

By the way, Article 4 of the EU Artificial Intelligence Act states:

“Providers and deployers of AI systems shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf, taking into account their technical knowledge, experience, education and training and the context the AI systems are to be used in, and considering the persons or groups of persons on whom the AI systems are to be used.”

Are you prepared?

References

  1. Bignami E, Russo M, Lanza R, Bellini V. Navigating the integration of large language models in healthcare: challenges, opportunities, and implications under the EU AI Act. J Anesth Analg Crit Care. 2024 Dec 2;4(1):79. doi: 10.1186/s44158-024-00215-w.
  1. Health Foundation. Priorities for an AI in health care strategy,https://www.health.org.uk/reports-and-analysis/briefings/priorities-for-an-ai-in-health-care-strategy, 2024
  1. Schmidt J, Schutte NM, Buttigieg S, Novillo-Ortiz D, Sutherland E, Anderson M, de Witte B, Peolsson M, Unim B, Pavlova M, Stern AD, Mossialos E, van Kessel R. Mapping the regulatory landscape for artificial intelligence in health within the European Union. NPJ Digit Med. 2024 Aug 27;7(1):229. doi: 10.1038/s41746-024-01221-6.
  1. Artificial Intelligence Act. Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU) No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives 2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) (Text with EEA relevance)http://data.europa.eu/eli/reg/2024/1689/oj
  1. Ho CW, Caals K. How the EU AI Act Seeks to Establish an Epistemic Environment of Trust. Asian Bioeth Rev. 2024 Jun 24;16(3):345-372. doi: 10.1007/s41649-024-00304-6.
  1. Gille F, Maaß L, Ho B, Srivastava D. From Theory to Practice: Viewpoint on Economic Indicators for Trust in Digital Health. J Med Internet Res 2025;27:e59111
    doi: 10.2196/59111

Author: Prof. Dr. Oliver Groene

Related Articles