Artificial intelligence (AI) systems are increasingly embedded in clinical diagnostics and referral pathways, yet when these tools contribute to patient harm, traditional medico-legal doctrines are strained by algorithmic opacity, automation bias, and distributed decision-making. Beyond legal uncertainty, AI integration raises fundamental ethical concerns regarding clinician moral agency, patient autonomy, and distributive justice. This scoping review maps how jurisdictions worldwide allocate legal and ethical responsibility for AI-assisted diagnostic and referral errors, identifies governance mechanisms that support accountability, and examines the implications for Saudi Arabia’s healthcare system under Vision 2030. Following the Arksey and O’Malley framework and PRISMA-ScR reporting guidelines, literature was identified through systematic searches of MEDLINE (via PubMed), Scopus, Web of Science, and CINAHL supplemented by targeted grey-literature retrieval from the World Health Organization, the European Commission, the U.S. Food and Drug Administration, and Gulf Cooperation Council regulatory sources. Sources were included if they substantively addressed legal liability, ethical accountability, or regulatory frameworks related to AI-assisted diagnostics or referrals. After screening and deduplication, 77 records were included. Data were extracted across eight thematic domains and analysed through both legal-doctrinal and ethical lenses informed by the principlist framework (autonomy, beneficence, non-maleficence, justice). Four dominant liability models were identified: negligence-based (27 studies), strict product liability (16), fault-based (12), and hybrid/tiered (11), with the remaining 11 records examining multiple or unspecified frameworks. Across all frameworks, clinicians bore disproportionate responsibility despite limited control over algorithmic processes, creating a “liability sink” phenomenon with significant ethical implications for moral distress and professional integrity. Only two studies directly addressed Saudi Arabia, revealing that 89% of clinicians lacked AI-related legal training, and consent forms failed to disclose AI involvement. Ethical analysis exposed tensions between innovation-driven policy goals and the principles of patient autonomy, non-maleficence, and justice. Effective AI governance is likely to depend on integrating legal and ethical frameworks. For Saudi Arabia, the limited jurisdiction-specific evidence supports the provisional development of a contextualised shared-responsibility model that could classify AI by clinical risk, draw on principlist ethical standards in regulatory design, address the moral dimensions of the liability sink, and align with Vision 2030’s innovation objectives and Islamic bioethical values. These tentative directions await empirical validation through Saudi-focused research.
Almutairi et al. (Mon,) studied this question.