EU adopts AI governance principles for health technology assessments

The EU’s Health Technology Assessment Coordination Group has adopted principles governing the use of AI in health technology assessments, requiring transparency, human oversight and accountability when AI assists in preparing evidence dossiers.

The guidance apply to dossiers submitted by health technology developers for Joint Clinical Assessments under the EU Health Technology Assessment Regulation. AI may be used to identify studies, extract data, assess the risk of bias, analyse evidence and prepare reports.

The group recognises that AI could improve the efficiency of evidence preparation but warns that inadequate oversight may compromise completeness, methodological quality and scientific rigour.

Health technology developers will remain fully responsible for the content, methods and conclusions of submitted dossiers, including decisions on whether AI is used, how it contributes to evidence synthesis and whether its outputs are scientifically valid.

Human oversight must be maintained throughout every AI-assisted stage, and no part of the assessment process should be fully automated without a clearly accountable individual.

Any AI-assisted step must be disclosed in the dossier, including information retrieval, study screening, data extraction, risk-of-bias assessment, evidence analysis and report preparation. Developers must also identify the AI tools used, including their name, version, developer and intended purpose.

Where commercial tools have been adapted, those modifications should be described, while prompts must be retained and made available during assessments on request.

The principles also require AI use to comply with applicable copyright, data protection and EU AI Act requirements, particularly given that submitted dossiers may later be published.

Why does it matter?

AI could accelerate the preparation and review of clinical evidence, but undisclosed automation or inaccurate outputs could influence decisions on which medicines and health technologies reach patients. By requiring human accountability, methodological transparency and disclosure of AI tools and prompts, the EU is seeking to preserve the scientific integrity of health technology assessments.

The principles also represent an early example of sector-specific AI governance under the EU’s broader regulatory framework. Rather than restricting AI use, they establish conditions for trustworthy deployment, illustrating how transparency and human oversight are becoming central requirements for AI-assisted decision-making in highly regulated fields.

Would you like to learn more about AI, tech, and digital diplomacy? If so, ask our Diplo chatbot!