Abstract The governance of artificial intelligence (AI) in European healthcare reveals a structural paradox. While the EU Artificial Intelligence Act (AI Act) classifies most medical AI systems as high-risk and imposes extensive obligations on providers and deployers, the institutional mechanisms meant to enforce these obligations often remain merely advisory. This article conceptualizes this disjunction as the enforceability gap: the systemic mismatch between formal legal responsibility and the institutional capacity to exercise coercive authority. This paper, in bridging insights from accountability theory, risk governance, and digital constitutionalism, argues that enforceability constitutes the missing link between algorithmic authority and democratic legitimacy. Authority without enforceability, it contends, results in symbolic governance, where ethics boards simulate oversight without power to constrain decisions. To overcome this deficit, the paper proposes a model of enforceable accountability grounded in four pillars: legally empowered institutional boards with conditional veto powers; corporate and professional liability mechanisms; transparency and restorative-oriented frameworks; and participatory procedures embedding democratic deliberation. The paper’s key innovation is to re-center hospitals as one-stop accountability hubs —integrated institutional loci where ethical review, legal compliance, and technical audit converge within a unified governance structure. By conceptualizing hospitals as active co-regulators rather than passive deployers, the paper demonstrates how enforceable accountability can be operationalized at the point of clinical implementation. In this perspective, addressing the enforceability gap is not solely a matter of regulatory refinement, but a structural condition for aligning high-risk AI governance with the rule-of-law requirements of the European digital order, ensuring that ethical obligations acquire effective legal force rather than remaining aspirational norms.
Maria Carla Canato (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: