Orhan Onder¹, Prana Rudra², Frank Ursin², Eike Buhr³

¹ History of Medicine and Ethics, School of Medicine, Marmara University

² Institute for Ethics, History and Philosophy of Medicine, Hannover Medical School

³ DivisionofEthicsin Medicine, Department of Health Services Research, University of Oldenburg

The integration of AI-based systems into medical care not only reshapes diagnostic and therapeutic practices but also alters the temporal structure of trust in clinical relationships. Whilst trust in medicine has traditionally developed over time through longitudinal interaction, professional judgement and shared vulnerability, AI-supported decision-making increasingly relies on asynchronous, opaque and often predictive processes. This raises the question of how trust – as distinct from mere reliance – can be sustained when clinical reasoning is partially delegated to systems whose reliability cannot be continuously or independently assessed.

Building on epistemological accounts that distinguish trust from mere reliance, we argue that AI-supported medical practice introduces a temporal tension between long-term trust relationships and short-term or retrospective evaluations of algorithmic performance. Whilst AI systems for prediction, risk stratification or clinical decision support may be underpinned by robust ex ante validation studies, their reliability in individual clinical encounters is typically only fully assessable ex post. By contrast, clinicians can be legitimately trusted ex ante on the basis of role-based responsibility, perceived goodwill, and relational cues such as empathy and responsiveness, even prior to demonstrable performance. Clinicians thus increasingly function as temporal proxies, expected to bridge reliability gaps over time by vouching for systems whose inner workings and future performance remain partially inaccessible to both professionals and patients.

We develop a conceptual analysis of this temporal asymmetry and examine its ethical implications for clinical judgement, responsibility, and the fiduciary role of clinicians. We argue that as AI systems become more autonomous within clinical practice, trust risks shifting away from shared clinical reasoning towards opaque forms of mediated reliance. Against this backdrop, we set out requirements for a temporally sensitive ethics of trust in AI-supported care, including the preservation of longitudinal clinical oversight, the integration of AI outputs into temporally informed diagnostic and therapeutic reasoning, and the clinician’s sustained responsibility for interpreting, contextualising and revising AI-mediated recommendations throughout the course of care.