>1,250
AI/ML medical devices already authorized by the FDA (mid-2025)[11]
variable
LLM quality and hallucination risk depend on context and sources[2]
Art. 4
EU AI Act: measures to support AI literacy, required since 02/02/2025, amended 2026[4]
7
EU HLEG principles of trustworthy AI[1]
What is it about?

Without a shared basic understanding, one of two things happens in daily practice: AI is either not used at all or used uncritically. Both cost impact. For AI to create real value in healthcare, organisations need a foundation training for employees with and without prior knowledge that is tailored to role, prior knowledge and context of use.

This is not only sound pedagogy; it has been a legal requirement since 2 February 2025. Article 4 of the AI Act obliges providers and deployers to take measures to support the development of AI literacy among their staff. Since the amendment by Regulation (EU) 2026/1744 (in force since 27 July 2026), no specific level of competence is prescribed; role, prior knowledge and context of use remain decisive.[4] The European Commission's questions and answers on AI literacy state explicitly that there is no one-size-fits-all approach and that simply passing on the instructions for use is not enough.[12]

An effective foundation training therefore combines technical principles (AI/ML/Deep Learning, algorithms, foundation models) with clinical practice (fields of application, metrics, prompting), ethical principles and the legal framework (EU AI Act, GDPR, MDR/IVDR). Central to this is the understanding that AI is not a neutral tool: systems can inherit bias from data, labels, measurement processes and organisational decisions.[1,4,6-9]

The content proposed here draws on peer-reviewed reviews, international guidance and applicable EU regulation.[1,4,7,9,13-15]

Needs-based, not one-size-fits-all

Core learning content in three tiers

Not everyone needs the same thing. Tier 1 applies to everyone, Tier 2 to everyone who works with AI outputs, Tier 3 to leadership and project ownership. The specific design follows from role, prior knowledge and the risk class of the systems in use.[4,12]

Tier 1

Foundations for all employees

Regardless of role and prior knowledge: the shared basis.

01

Basic terms: AI, ML & Deep Learning

Differentiation of terms, ML learning types (supervised/unsupervised/reinforcement), classical algorithms (decision trees to neural networks), foundation models and LLMs. Strengths and limitations in the clinical context.[2,7,9]

02

What AI can do and what it cannot

Fields of application from radiology and imaging through decision support, early detection and automatic coding to monitoring and genomics, each set against error risks, bias, validation limits and the hallucination risk of generative systems.[2,3,7,9]

03

Safe in daily work: data, purpose limitation, shadow AI

What must not go into a freely accessible tool, purpose limitation and legal bases under the GDPR for health data, handling of unapproved tools and clear escalation routes in case of doubt.[6,9]

04

Transparency: knowing the obligations, making AI contributions traceable

Transparency obligations under Art. 50 of the EU AI Act (e.g. notice when people interact directly with AI systems, labelling of synthetic content), informing patients under the GDPR and medical duties to inform, and documented separation between AI suggestion and human decision.[4,5]

Tier 2

Depth for employees who work with AI outputs

Anyone who assesses AI outputs or lets them feed into decisions needs judgement, not just operating knowledge.

05

Metrics: sensitivity, specificity, prevalence

Trade-off between sensitivity and specificity, the importance of prevalence (base-rate problem), calibration and why models can perform significantly worse outside their training and validation context.[7,10]

06

Bias & fairness: why AI can discriminate

Representation bias, historical bias, measurement bias, label bias and subgroup performance. XAI methods such as SHAP or Grad-CAM can support analysis and communication, but they do not replace validation, fairness testing and clinical assessment.[7,8]

07

Human oversight in practice

Human-in-the-loop in concrete terms: recognising automation bias, being able to contradict the system with reasons, and keeping professional responsibility clearly assigned. Oversight must be defined operationally, not just on paper.[1,4,9]

08

Prompting & documentation

Effective prompting (context, role, requesting sources, addressing uncertainty), critical review of generative outputs and separate documentation of AI recommendations and human decisions.[2,4,9]

Tier 3

Leadership, governance and project ownership

Anyone deciding on adoption, procurement and oversight needs both the regulatory and the strategic frame.

09

Ethics & law at a glance

7 EU HLEG requirements for trustworthy AI,[1] the EU AI Act with its risk-based approach, prohibited practices, high-risk obligations and transparency requirements,[4,5] GDPR for personal health data and MDR/IVDR when there is an intended medical purpose.[3,4,6] In addition, the European Health Data Space (EHDS, Regulation (EU) 2025/327) addresses the use of health data.

10

Target vision and use-case decisions

Four design principles (human centredness, traceability, relief, distributed control) and the three key questions before any deployment. Plus go / wait / stop as decision logic, allocation of liability and measurable KPIs for adoption.[1,7,9]

Classification

For whom and why

This outline addresses a practical gap in many healthcare organizations: employees who work with AI outputs or accompany AI implementation projects need basic knowledge of data quality, validation, bias, human oversight and regulatory classification. Without this knowledge, the strengths and limitations of real AI systems are difficult to assess, with consequences for patient safety, compliance and trust.[1,7,9]

The gap is documented: reviews identify the lack of standardised AI education as a key barrier to adoption and note at the same time that consistent curriculum frameworks are still missing and need to be adaptable.[14,15] Targeted training substantially improves AI knowledge and the effect is largely maintained; role-specific adoption gaps are visible in the evidence, which is a further argument against a single format for everyone.[13]

The content follows the European regulatory framework.

trainingAI BasicsEU AI Act GDPRBias & fairnessHealth Literacy AI literacy (Art. 4)Human oversight
Make a request →
Further reading Humane AI use & vision How to introduce AI in a human-centred way: four design principles, three key questions and practical healthcare examples.
To the vision →
Bibliography

Sources

Unless otherwise stated, retrieved: April 2026; sources [12]–[15] retrieved in August 2026.

  1. High-Level Expert Group on Artificial Intelligence. Ethics guidelines for trustworthy AI. Brussels: European Commission; 2019.
  2. Iqbal U, Tanweer A, Rahmanti AR, Greenfield D, Lee LTJ, Li YCJ. Impact of large language model (ChatGPT) in healthcare: an umbrella review and evidence synthesis. J Biomed Sci. 2025;32:45. doi:10.1186/s12929-025-01131-z.
  3. Aboy M, Minssen T, Vayena E. Navigating the EU AI Act: implications for regulated digital medical products. NPJ Digit Med. 2024;7:237. doi:10.1038/s41746-024-01232-3.
  4. European Parliament, Council of the European Union. Regulation (EU) 2024/1689 of 13 June 2024 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act). Official Journal of the European Union. 2024 Jul 12;L2024/1689. Official text (ELI): eur-lex.europa.eu/eli/reg/2024/1689/oj.
    Art. 4: AI literacy (applicable since 02/02/2025, Art. 113(a); as amended by Regulation (EU) 2026/1744, in force since 27/07/2026): providers and deployers of AI systems take measures to support the development of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf; a specific level of AI literacy is no longer required. Amending regulation: eur-lex.europa.eu/eli/reg/2026/1744/oj. See also Art. 50 (transparency obligations).
  5. Gilbert S. The EU passes the AI Act and its implications for digital medicine are unclear. NPJ Digit Med. 2024;7:135. doi:10.1038/s41746-024-01116-6.
  6. European Parliament, Council of the European Union. Regulation (EU) 2016/679 of 27 April 2016 on the protection of natural persons with regard to processing of personal data and on the free movement of such data. Official Journal of the European Union. 2016 May 4;L119:1-88.
  7. Lekadir K, Quaglio G, Tselioudis Garmendia A, Gallin C. Artificial intelligence in healthcare: applications, risks, and ethical and societal impacts. Brussels: European Parliamentary Research Service; 2022. Available from: https://www.europarl.europa.eu/thinktank/en/document/EPRS_STU(2022)729512.
  8. Yang Y, Lin M, Zhao H, Peng Y, Huang F, Lu Z. A survey of recent methods for addressing AI fairness and bias in biomedicine. J Biomed Inform. 2024;154:104646. doi:10.1016/j.jbi.2024.104646.
  9. World Health Organization. Regulatory considerations on artificial intelligence for health. Geneva: World Health Organization; 2023.
  10. Collins GS, Reitsma JB, Altman DG, Moons KGM. Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162:55-63.
  11. U.S. Food and Drug Administration. Artificial Intelligence-Enabled Medical Devices. Silver Spring (MD): FDA; 2025 (as of mid-2025: >1,250 authorized products; accessed June 2026). Available from: fda.gov. Peer-reviewed analysis: Singh R, Bapna M, Diab AR, Ruiz ES, Lotter W. How AI is used in FDA-authorized medical devices: a taxonomy across 1,016 authorizations. NPJ Digit Med. 2025. doi:10.1038/s41746-025-01800-1.
  12. European Commission. AI Literacy – Questions & Answers. Brussels: European Commission; continuously updated document, as of 27/07/2026. Available from: digital-strategy.ec.europa.eu.
  13. El Arab RA, Alshakihs AH, Alabdulwahab SH, Almubarak YS, Alkhalifah SS, Abdrbo A, Hassanein S, Sagbakken M. Artificial intelligence in nursing: a systematic review of attitudes, literacy, readiness, and adoption intentions among nursing students and practicing nurses. Front Digit Health. 2025;7:1666005. doi:10.3389/fdgth.2025.1666005.
  14. El Arab RA, Al Moosa OA, Abuadas FH, Somerville J. The role of AI in nursing education and practice: umbrella review. J Med Internet Res. 2025;27:e69881. doi:10.2196/69881.
  15. Hernández Rincón EH, Jimenez D, Chavarro Aguilar LA, Pérez Flórez JM, Romero Tapia ÁE, Jaimes Peñuela CL. Mapping the use of artificial intelligence in medical education: a scoping review. BMC Med Educ. 2025;25:526. doi:10.1186/s12909-025-07089-8.

Note on the evidence base: reviews [13]–[15] come predominantly from nursing and medical education settings. They support the effectiveness of training and the existence of role-specific adoption gaps; transfer to industry and administrative settings is plausible but not separately evidenced.

⚠ Disclaimer

This material was created with the greatest possible care. It is not a substitute for legal, medical or professional advice. References to the EU AI Act, GDPR and MDR/IVDR are provided for orientation and do not constitute legal advice. No guarantee for completeness or topicality.

Note: AI-assisted creation (voluntary notice in line with EU AI Act Art. 50)

Parts of this document were created and editorially checked with the support of generative AI. According to EU AI Act Art. 50, the use of AI is transparently pointed out.