Analysis of Generative Artificial Intelligence Governance in Publicly Available Documentation of Global and Saudi Universities Offering Health Informatics Programs Authors: Anas Ali Alhur, Nouf Khalid Al-Kahtani Published: 2026 ABSTRACT BackgroundGenerative artificial intelligence (GenAI) has rapidly entered higher education, prompting universities to publish policies and guidance for its responsible use. Little is known about how GenAI is governed within Health Informatics and related programs, or how international and Saudi institutions compare.ObjectiveTo analyze and compare publicly available GenAI governance in global and Saudi universities offering Health Informatics and related programs, focusing on policy availability, governance-domain coverage, and gaps.MethodsA qualitative document analysis and comparative policy analysis were conducted. Thirty international universities (selected via the QS World University Rankings by Subject 2025) and sixteen eligible Saudi institutions offering Health Informatics, Health Information Management, Digital Health, or related programs were included. Publicly available governance sources—comprising formal policy documents and web-based institutional guidance—were retrieved from official institutional websites between March and May 2026 and analyzed using directed qualitative content analysis across nine governance domains. Inter-coder reliability was almost perfect (Cohen’s κ = 0.96), and group comparisons were tested using Fisher’s exact test.ResultsIn total, 117 governance sources met the inclusion criteria (111 international; 6 Saudi). All international universities (30/30) had publicly available GenAI governance, versus six of sixteen Saudi universities (37.5%). International coverage was significantly greater for Transparency and Disclosure (87.4% vs 33.3%; p=0.005), Ethical Considerations (91.9% vs 50.0%; p=0.014), and AI Literacy and Capacity Building (81.1% vs 33.3%; p=0.019). Program-specific Health Informatics governance appeared at only two international institutions and no Saudi universities; clinical-data-governance provisions were absent from all Saudi sources.ConclusionBased on publicly available documentation, international universities showed broader and deeper GenAI governance than Saudi institutions, with gaps in transparency, AI literacy, ethics, and Health Informatics-specific and clinical-data governance. Strengthening discipline-specific guidance may support responsible GenAI integration and digital-health workforce preparation. Introduction Generative artificial intelligence (GenAI) has attracted rapid and widespread attention across higher education. These technologies rely on large language models such as ChatGPT, which are trained on massive datasets and can generate human-like text, code, and images.1 Their rapid expansion has created new opportunities for teaching, learning, research, and assessment, while simultaneously raising concerns regarding academic integrity, accuracy, transparency, and responsible use.2 In response, universities worldwide have begun publishing policies, guidance, and governance frameworks to shape how GenAI is used by students and faculty.3–5 GenAI is increasingly used across health professions education, including medicine, nursing, pharmacy, and allied disciplines, where it can support information retrieval, self-directed learning, academic writing, and access to educational resources.6–12 Concerns regarding accuracy, reliability, bias, and overreliance are especially relevant when GenAI is used for learning and decision support.13 Evidence from education and health-professions education has documented both the educational potential of GenAI and concerns regarding its appropriate and responsible use, including implications for medical and nursing education and the risk of dependency.14–17 Health Informatics differs from many traditional health professions because it combines healthcare knowledge with information technology, health information systems, data analytics, and digital health technologies.18 Graduates are likely to work directly with AI-enabled healthcare systems, protected health data, and clinical decision-support tools. Across the broader GenAI literature, students and educators have reported potential benefits alongside concerns about model capabilities, limitations, and wider educational and societal implications.19–21 Academic-integrity concerns have also intensified as GenAI has become more accessible to students.22 Systematic and conceptual work in higher education further highlights the need for responsible institutional approaches that balance educational opportunities with ethical and academic considerations.23–25 These governance questions are particularly relevant to assessment, where GenAI has challenged conventional approaches to authorship, originality, academic integrity, and assessment design.26 For clarity, this study distinguishes four related but conceptually different elements of institutional response. Governance frameworks refer to the overarching principles and structures that orient an institution’s approach to GenAI. Institutional policies are formal rules specifying permitted, restricted, and prohibited uses. Implementation mechanisms are the operational procedures—training, disclosure workflows, approved platforms—through which policies are enacted. Accountability structures define oversight, monitoring, and enforcement. Throughout, we use the umbrella term GenAI governance to encompass these elements as they appear in publicly available institutional documentation. It is important to state at the outset that this study examines only publicly accessible institutional documentation. Governance that institutions maintain internally—but do not publish—was outside the scope of what could be observed, and the availability (or absence) of a public document is therefore treated as a finding about transparency rather than as definitive evidence of the presence or absence of internal governance. Recent international developments have sharpened the need for such analysis. WHO guidance on the ethics and governance of AI for health emphasizes privacy, accountability, human oversight, and patient protection across the AI lifecycle.27 Emerging evidence has also raised concerns about the potential effects of AI-tool use on critical thinking, reinforcing the importance of governance that supports independent reasoning and critical appraisal.28 Complementary guidance from the European Commission and university-sector principles has emphasized ethical and responsible AI use and the need to support students and staff in using GenAI appropriately.29,30 In the Saudi context, interest in AI is evident within health-sciences education, including research examining university students’ attitudes toward AI in healthcare.31 Recent evaluations in high-stakes health-professions contexts further illustrate the need for critical appraisal: studies of generative AI performance in dental licensing examinations and chatbot responses to clinical multiple-choice questions show that outputs may be useful yet inconsistent or inaccurate.32,33 Prior qualitative work by the present authors has also explored student engagement, perceived educational implications, and ethical considerations of GenAI in Health Informatics education, providing discipline-specific grounding for the governance focus adopted here.34 Several gaps therefore remain. Few studies have systematically compared GenAI governance across multiple universities; little attention has been directed to Health Informatics education despite its close relationship with AI and health data; and there is limited evidence regarding GenAI governance in Saudi universities relative to leading international institutions. This study addresses these gaps through a comparative analysis of publicly available GenAI governance. Study Aim To analyze and compare publicly available generative artificial intelligence governance in global and Saudi universities offering Health Informatics and related programs, with a focus on institutional governance approaches, policy content, and governance gaps. Study Objectives To identify and analyze publicly available policies and guidance related to GenAI in selected global and Saudi universities offering Health Informatics and related programs.To compare the content and governance approaches of these sources, including acceptable use, academic integrity, ethics, privacy, transparency, and implementation mechanisms.To identify governance gaps and provide recommendations for strengthening GenAI governance within Health Informatics education. To identify and analyze publicly available policies and guidance related to GenAI in selected global and Saudi universities offering Health Informatics and related programs. To identify and analyze publicly available policies and guidance related to GenAI in selected global and Saudi universities offering Health Informatics and related programs. To compare the content and governance approaches of these sources, including acceptable use, academic integrity, ethics, privacy, transparency, and implementation mechanisms. To compare the content and governance approaches of these sources, including acceptable use, academic integrity, ethics, privacy, transparency, and implementation mechanisms. To identify governance gaps and provide recommendations for strengthening GenAI governance within Health Informatics education. To identify governance gaps and provide recommendations for strengthening GenAI governance within Health Informatics education. Research Questions What publicly available GenAI policies and guidance are currently available in selected global and Saudi universities offering Health Informatics and related programs?How do these sources differ in their governance approaches to GenAI use in higher education?What governance gaps can be identified, and how might institutional governance of GenAI be strengthened within Health Informatics education? What publicly available GenAI policies and guidance are currently available in selected global and Saudi universities offering Health Informatics and related programs? What publicly available GenAI policies and guidance are currently available in selected global and Saudi universities offering Health Informatics and related programs? How do these sources differ in their governance approaches to GenAI use in higher education? How do these sources differ in their governance approaches to GenAI use in higher education? What governance gaps can be identified, and how might institutional governance of GenAI be strengthened within Health Informatics education? What governance gaps can be identified, and how might institutional governance of GenAI be strengthened within Health Informatics education? Methods Study Design This study employed a qualitative document analysis and comparative policy analysis design to examine publicly available institutional governance of GenAI in higher education institutions offering Health Informatics and related programs. Document analysis was selected because publicly available institutional documentation represents the formal, observable expression of an institution’s governance stance—its stated expectations, responsibilities, and regulatory mechanisms. The unit of analysis was the publicly available governance source; public availability itself was treated as a substantive dimension of the phenomenon under study, because the transparency with which institutions publish AI governance is part of what this study sought to characterize. A governance source was defined as a discrete, publicly available institutional artifact addressing GenAI, comprising either a formal policy document (eg, a downloadable PDF policy, handbook, or regulation) or web-based institutional guidance (eg, a policy or guidance page published on an official university website). Both forms were treated as units of analysis because institutions publish GenAI governance in both formats; the majority of sources identified were web-based institutional guidance rather than formal downloadable documents, and the composition by format is reported in the Results. The study was informed by directed qualitative content analysis, which permits examination of documents against predefined analytical categories while remaining open to emergent themes. This approach suited a domain in which international guidance provided an initial deductive basis for coding, while the rapidly evolving nature of GenAI governance required openness to newly emerging policy domains. University Selection International Universities International universities were selected using the QS World University Rankings by Subject, 2025 edition. Because QS does not provide a dedicated subject ranking for Health or Medical Informatics, institutions were identified from the four subject tables most closely associated with the discipline: Medicine; Life Sciences and Medicine; Computer Science and Information Systems; and Data Science and Artificial Intelligence. Institutions appearing in more than one subject table were de-duplicated and counted once. Universities were eligible if they (i) appeared within the ranked institutions of one or more of these subject tables in the 2025 edition and (ii) offered a formal academic program, department, institute, center, or structured pathway related to Health Informatics, Medical Informatics, Biomedical Informatics, Clinical Informatics, Digital Health, Health Information Management, Health Information Technology, Health Data Science, or a closely related discipline. Applying these criteria yielded 30 international universities from North America, Europe, Asia, and Australia. Details of the included institutions, their Health Informatics-related academic programs, and governance approaches are provided in Supplementary Tables S1 and S2. Saudi Universities Saudi institutions were eligible if they officially offered a formal academic program, department, or specialization in Health Informatics, Health Information Management, Health Information Technology, Health Information Management and Technology, or a closely related discipline. Institutions were identified through the Ministry of Education program directory, official institutional websites, academic program directories, and publicly available program information, using both English and Arabic search terms Table S1. Sixteen Saudi institutions met the eligibility criteria and were included. The selection process at both institution and source levels is summarized in the PRISMA-style flow diagram (Figure 1). Figure 1Temporal Distribution of Publicly Available Generative Artificial Intelligence Governance Documents by Publication Year and University Group.A grouped bar graph showing counts of governance documents by publication year for international and Saudi groups. Temporal Distribution of Publicly Available Generative Artificial Intelligence Governance Documents by Publication Year and University Group. Source Identification and Search Strategy Publicly available governance sources were retrieved from official institutional channels between March and May 2026, including university websites, academic-integrity offices, teaching and learning centers, assessment and examination policies, AI guidance pages, digital-learning units, student and faculty handbooks, institutional governance frameworks, and other official policy repositories. Both university-wide and faculty- or program-specific sources were reviewed, whether directed toward students, faculty, or both. When a university had no standalone GenAI policy, broader institutional documents were included if they explicitly addressed artificial intelligence, GenAI, ChatGPT, large language models, or AI-assisted academic work. Systematic institutional-site searches were supplemented by external search-engine queries using terms including “generative AI policy”, “artificial intelligence policy”, “ChatGPT policy”, “AI guidance”, “AI governance”, “AI use in assessment”, “academic integrity and AI”, “responsible AI use”, and “large language model policy”, individually and in combination, with equivalent Arabic terms for Saudi institutions. All retrieved sources were downloaded or archived and catalogued in a searchable database. Inclusion and Exclusion Criteria Sources were included if they were published or endorsed by an official institutional source; explicitly referred to GenAI, artificial intelligence, ChatGPT, large language models, or related technologies; addressed educational, academic, teaching, learning, assessment, research, or governance contexts; applied to students, faculty, or academic staff; were publicly accessible during data collection; and were available in English or Arabic. Sources were excluded if they addressed only technical IT infrastructure; focused exclusively on cybersecurity without reference to educational AI use; were duplicate versions of existing policies; were unofficial blogs, opinion pieces, news articles, or promotional materials; or contained no reference to artificial intelligence or GenAI. Analytical Framework and Domain Development A nine-domain coding framework was developed deductively from established international guidance and refined inductively. The domains were derived from UNESCO guidance for generative AI in education and research, WHO guidance on the ethics and governance of AI for health, and the OECD AI Principles, supplemented by higher-education literature on educational-technology governance, AI ethics, and academic integrity. Each domain maps to specific source guidance (Supplementary Tables S3): Academic Integrity; Acceptable and Prohibited Use; Assessment Governance; Transparency and Disclosure; Ethical Considerations; Privacy and Data Protection; AI Literacy and Capacity Building; Governance and Accountability; and Cognitive and Educational Considerations. Domain definitions used as the coding standard are provided in the Supplementary Material. Coding, Reliability, and Analysis All sources were imported into NVivo for qualitative analysis. Coding proceeded in two stages: deductive coding, in which content was assigned to one or more of the predefined domains, followed by inductive coding, in which emergent themes not captured by the framework were incorporated. A standardized extraction form (university, country, program classification, source title, year, source type, audience, scope, domains addressed, key provisions, implementation mechanisms, and enforcement) ensured consistency across institutions. Coding reliability was formally assessed. A second coder independently coded the international corpus (n = 91 sources) across the nine governance domains, following shared written domain definitions and a common codebook, with coder training conducted before independent coding. Inter-coder agreement was almost perfect (Cohen’s κ = 0.96; 98.9% raw agreement across 819 paired judgments). Per-domain agreement exceeded 98% for all nine domains; the lower κ observed for Acceptable and Prohibited Use (κ = 0.49) reflects the well-recognized kappa base-rate paradox arising from that domain’s near-universal presence (>98% of sources) rather than genuine coder disagreement. Remaining discrepancies were resolved through discussion to consensus, and the reconciled scheme was applied to the full corpus. NVivo coding procedures were standardized through the shared codebook and periodic consistency checks. Comparative analysis was conducted between international and Saudi universities across policy scope, governance priorities, implementation approaches, ethical considerations, transparency requirements, academic-integrity provisions, and cognitive concerns. Descriptive statistics summarized policy availability and governance-domain frequencies at both the source level and the institution level. Because the two groups differed markedly in the number of available sources (111 vs 6), raw counts are reported alongside all percentages, and small-denominator percentages (Saudi group) are interpreted as descriptive rather than precise magnitude estimates. To test group differences while respecting these small, unequal cell sizes, two-sided Fisher’s exact tests were applied to each domain comparison; p <0.05 was considered statistically significant. Content Intensity and Governance Maturity Beyond binary domain presence, a content-intensity analysis assessed governance depth using four ordered levels: Absent (domain not addressed), Present (surface-level mention only), Operational (specific, actionable provisions stated), and Embedded (detailed, multi-provision coverage integrated across the source). A composite governance-maturity score (0–100) was then computed from five weighted components—Domain Coverage, Document Depth, Verbatim Evidence, Health Informatics Specificity, and Governance Orientation—with the scoring rules, maxima, and thresholds specified in Supplementary Table S4. This composite framework was developed specifically for the present study, drawing conceptually on maturity-model approaches used in digital and health-informatics governance; it has not been externally validated, a limitation noted below. Because the assessment is based on publicly available documentation, “maturity” throughout denotes the maturity of publicly available governance documentation rather than an institution’s complete internal governance. Trustworthiness, Reflexivity, and Ethics Credibility was supported by the use of official institutional sources and systematic retrieval; dependability by a predefined search strategy, standardized extraction, and documented coding; confirmability by a complete audit trail of sources and coding decisions; and transferability by the inclusion of geographically diverse institutions. Reflexively, the authors acknowledge that their disciplinary background in Health Informatics shaped the analytical emphasis on clinical-data and discipline-specific governance; coding definitions and consensus procedures were used to mitigate individual interpretive bias. The study analyzed only publicly available institutional documents and did not involve human participants, personal information, or identifiable data; formal ethical approval was therefore not required. The study adhered to principles of responsible research conduct, transparency, and accurate representation of institutional documentation. Results Policy Availability and Source Characteristics A systematic search of official institutional websites identified 117 publicly available governance sources across the 46 universities that met the inclusion criteria (111 international; 6 Saudi). All 30 international universities (100.0%) had at least one publicly available GenAI governance source, compared with six of 16 Saudi universities (37.5%); ten Saudi institutions (62.5%) had no publicly available governance documentation at the time of data collection (Table 1). Consistent with the definition adopted here, the majority of international sources were web-based institutional guidance rather than formal downloadable documents, reflecting how institutions predominantly publish GenAI governance.Table 1Policy Availability and Source Characteristics by University GroupCharacteristicInternational Universities (n=30)Saudi Universities (n=16)Universities with ≥1 publicly available policy document30 (100.0%)6 (37.5%)Universities with no publicly available policy document0 (0.0%)10 (62.5%)Total substantive documents identified1116Mean documents per university (range)3.7 (1–12)0.4 (0–1)Institutions with ≥5 documents (multi-layer governance)5 (16.7%)0 (0.0%)Documents published 2023–202699 (89.2%)4 (66.7%)Documents addressing both students and faculty94 (84.7%)4 (66.7%)Documents with explicit health informatics provisions50 (45.0%)1 (16.7%)Notes: Health-informatics provisions are defined as explicit reference to clinical data governance, HIPAA/PHI, patient-data protection, or program-specific AI-competency requirements. Multi-layer governance = an institution with sources at ≥2 organizational levels (eg, university-wide + school-level + program-level). Table 1Policy Availability and Source Characteristics by University GroupCharacteristicInternational Universities (n=30)Saudi Universities (n=16)Universities with ≥1 publicly available policy document30 (100.0%)6 (37.5%)Universities with no publicly available policy document0 (0.0%)10 (62.5%)Total substantive documents identified1116Mean documents per university (range)3.7 (1–12)0.4 (0–1)Institutions with ≥5 documents (multi-layer governance)5 (16.7%)0 (0.0%)Documents published 2023–202699 (89.2%)4 (66.7%)Documents addressing both students and faculty94 (84.7%)4 (66.7%)Documents with explicit health informatics provisions50 (45.0%)1 (16.7%)Notes: Health-informatics provisions are defined as explicit reference to clinical data governance, HIPAA/PHI, patient-data protection, or program-specific AI-competency requirements. Multi-layer governance = an institution with sources at ≥2 organizational levels (eg, university-wide + school-level + program-level). Policy Availability and Source Characteristics by University Group Notes: Health-informatics provisions are defined as explicit reference to clinical data governance, HIPAA/PHI, patient-data protection, or program-specific AI-competency requirements. Multi-layer governance = an institution with sources at ≥2 organizational levels (eg, university-wide + school-level + program-level). Temporal Distribution Among international universities, most sources were published between 2023 and 2026 (Figure 1), reflecting the institutional response that followed the public release of ChatGPT in late 2022. Governance activity among Saudi universities was limited across the same period. Annual source counts for each group are shown in Figure 1 and reconcile to the group totals (111 international; 6 Saudi). For the international group, governance sources were distributed as 2023 (n = 43), 2024 (n = 20), 2025 (n = 29), 2026 (n = 7), and undated (n = 12), summing to 111; the six Saudi sources comprised two published before 2022, two in 2023, one in 2024, and one in 2026 (Figure 1). Governance-Domain Coverage At the source level, international institutions demonstrated higher coverage across seven of the nine governance domains (Table 2 and Supplementary Figure S1), a pattern that is also evident in the university-level governance-domain coverage heat map (Figure 2). The largest differences were observed for Transparency and Disclosure (+54.1 percentage points), AI Literacy and Capacity Building (+47.7), and Ethical Considerations (+41.9). Acceptable and Prohibited Use and Governance and Accountability were comparable between groups (differences < 2 percentage points).Table 2Governance-Domain Coverage by University Group (Source-Level), with Fisher’s Exact TestsGovernance DomainInt’l n/NInt’l %Saudi n/NSaudi %Diff. (pp)PatternFisher pAcademic Integrity96/11186.5%4/666.7%+19.8International0.210Acceptable & Prohibited Use110/11199.1%6/6100.0%−0.9Comparable1.000Assessment Governance80/11172.1%4/666.7%+5.4Comparable1.000Transparency & Disclosure97/11187.4%2/633.3%+54.1International0.005**Ethical Considerations102/11191.9%3/650.0%+41.9International0.014*Privacy & Data Protection77/11169.4%3/650.0%+19.4International0.379AI Literacy & Capacity Building90/11181.1%2/633.3%+47.7International0.019*Governance & Accountability109/11198.2%6/6100.0%−1.8Comparable1.000Cognitive & Educational Considerations87/11178.4%4/666.7%+11.7International0.613Notes: N = total sources per group (International = 111; Saudi = 6). pp = percentage points. Diff. = International % − Saudi %. Two-sided Fisher’s exact test: * p <0.05; ** p <0.01. Cells marked “comparable” differ by < 2 pp. Figure 2Governance-domain coverage heat map for all 46 universities across the nine analytical domains.A large grid compares 46 universities (listed in rows) against nine governance-related policy areas (shown as column headers). Each cell contains a checkmark and is color-coded: green indicates the institution addresses that area in at least one document, while red indicates it is not addressed. A rightmost column summarizes each university with a domains covered total and a bottom row shows overall coverage percentages per domain. The upper portion of the table is dominated by green cells (broad coverage), while the lower portion contains many red cells (limited or no coverage), creating a clear contrast between higher- and lower-coverage institutions.Matrix of 46 universities with green/red checkmarks across nine policy areas and a total score column. Table 2Governance-Domain Coverage by University Group (Source-Level), with Fisher’s Exact TestsGovernance DomainInt’l n/NInt’l %Saudi n/NSaudi %Diff. (pp)PatternFisher pAcademic Integrity96/11186.5%4/666.7%+19.8International0.210Acceptable & Prohibited Use110/11199.1%6/6100.0%−0.9Comparable1.000Assessment Governance80/11172.1%4/666.7%+5.4Comparable1.000Transparency & Disclosure97/11187.4%2/633.3%+54.1International0.005**Ethical Considerations102/11191.9%3/650.0%+41.9International0.014*Privacy & Data Protection77/11169.4%3/650.0%+19.4International0.379AI Literacy & Capacity Building90/11181.1%2/633.3%+47.7International0.019*Governance & Accountability109/11198.2%6/6100.0%−1.8Comparable1.000Cognitive & Educational Considerations87/11178.4%4/666.7%+11.7International0.613Notes: N = total sources per group (International = 111; Saudi = 6). pp = percentage points. Diff. = International % − Saudi %. Two-sided Fisher’s exact test: * p <0.05; ** p <0.01. Cells marked “comparable” differ by < 2 pp. Governance-Domain Coverage by University Group (Source-Level), with Fisher’s Exact Tests Notes: N = total sources per group (International = 111; Saudi = 6). pp = percentage points. Diff. = International % − Saudi %. Two-sided Fisher’s exact test: * p <0.05; ** p <0.01. Cells marked “comparable” differ by < 2 pp. Governance-domain coverage heat map for all 46 universities across the nine analytical domains. Because the groups differed markedly in the number of available sources, two-sided Fisher’s exact tests were applied to each domain comparison, with raw counts reported alongside percentages. The three principal contrasts remained statistically significant despite the small Saudi sample: Transparency and Disclosure (p =0.005), Ethical Considerations (p =0.014), and AI Literacy and Capacity Building (p =0.019). Domains described as comparable were non-significant, indicating that the observed divergences are not artifacts of unequal denominators. University-Level Coverage At the institution level, the governance-domain coverage heat map (Figure 2) demonstrates that all 30 international universities addressed eight or more of the nine governance domains through their publicly available documentation (mean 8.8 domains). In contrast, among Saudi universities, two institutions addressed eight domains, four addressed five to seven domains, and ten had no publicly available governance documentation (mean 2.1 domains). Reporting at both the source and institution levels (Tables 2) shows the same directional pattern, and the overall university-level governance classification is presented in (Table 3).Table 3University-Level Governance-Coverage ClassificationCategoryDefining CharacteristicsInt’l n (%)Saudi n (%)Saudi Institutional ExamplesComprehensiveMulti-document portfolio; enabling and compliance provisions; ≥8 of 9 domains addressed30 (100%)2 (12.5%)King Abdulaziz University (KAU); Imam Abdulrahman Bin Faisal University (IAU)ModerateSingle compliance document; basic provisions across 5–7 domains0 (0.0%)4 (25.0%)KSAU-HS; Saudi Electronic University; Majmaah University; King Faisal UniversityNoneNo publicly available governance documentation identified0 (0.0%)10 (62.5%)University of Hail, Qassim University, Jazan University, Umm Al-Qura University, and six othersNotes: Comprehensive = ≥8 of 9 domains addressed. None = no publicly available governance documentation identified. No “Limited” category (1–4 domains) was observed in either group. Table 3University-Level Governance-Coverage ClassificationCategoryDefining CharacteristicsInt’l n (%)Saudi n (%)Saudi Institutional ExamplesComprehensiveMulti-document portfolio; enabling and compliance provisions; ≥8 of 9 domains addressed30 (100%)2 (12.5%)King Abdulaziz University (KAU); Imam Abdulrahman Bin Faisal University (IAU)ModerateSingle compliance document; basic provisions across 5–7 domains0 (0.0%)4 (25.0%)KSAU-HS; Saudi Electronic University; Majmaah University; King Faisal UniversityNoneNo publicly available governance documentation identified0 (0.0%)10 (62.5%)University of Hail, Qassim University, Jazan University, Umm Al-Qura University, and six othersNotes: Comprehensive = ≥8 of 9 domains addressed. None = no publicly available governance documentation identified. No “Limited” category (1–4 domains) was observed in either group. University-Level Governance-Coverage Classification Notes: Comprehensive = ≥8 of 9 domains addressed. None = no publicly available governance documentation identified. No “Limited” category (1–4 domains) was observed in either group. Content Intensity and Governance Maturity Content-intensity analysis (Figure 3) showed that international universities were predominantly at Operational and Embedded levels across most domains, whereas Saudi universities were most frequently at the Present level, with Operational-level coverage concentrated at Imam Abdulrahman Bin Faisal University and King Abdulaziz University. Composite governance-maturity scores (Figure S2; scoring rules in Table S2) were correspondingly higher among international institutions. These maturity results describe the maturity of publicly available documentation and should not be read as evaluations of complete internal governance. Figure 3Content-intensity heat map showing depth of governance provision (Absent, Present, Operational, Embedded) per institution and domain.The table outlines various universities and their approach to AI-related topics across multiple columns: Academic Integrity, Acceptable/Prohibited Use, Assessment, Transparency and Disclosure, Ethical Use, Privacy and Data Protection, AI Literacy and Capacity Building, Governance and Accountability, Cognitive and Educational Guidance. Most universities, including Harvard, Stanford, Cambridge, JHU, Imperial, UCL, Toronto, McGill, Yale, Columbia, UPenn, Michigan, UW, Pittsburgh, Vanderbilt, NUS, Sydney, Melbourne, Manchester, King′s, Karolinska, HKU and Cornell, have ′Embedded′ provisions across all columns. Oxford and Edinburgh have ′Operational′ provisions for Academic Integrity. UCSF has ′Present′ provisions across all columns. Duke lacks Transparency and Disclosure, while UCSD lacks Transparency and Disclosure and Ethical Use. Saudi universities show varied levels of provision, including Embedded, Present, Operational and Absent.A table comparing governance provision levels across international and Saudi universities in various domains. Content-intensity heat map showing depth of governance provision (Absent, Present, Operational, Embedded) per institution and domain. Comparable Domains Acceptable and Prohibited Use was identified in 110 of 111 international sources (99.1%) and all six Saudi sources (100.0%); Governance and Accountability in 109 of 111 (98.2%) and all six (100.0%). Although coverage was comparable, the provisions differed qualitatively: Saudi sources emphasized sanctions, disciplinary procedures, monitoring, and compliance, whereas international sources more often added human oversight of AI-generated outputs, faculty responsibilities, student disclosure, and institutionally approved AI platforms. High-Divergence Domains Transparency and Disclosure appeared in 97 of 111 international sources (87.4%) and two of six Saudi sources (33.3%; p =0.005), commonly including disclosure of AI use, citation of AI tools, and acknowledgement of AI-assisted work. AI Literacy and Capacity Building appeared in 90 of 111 (81.1%) and two of six (33.3%; p =0.019), including training, educational resources, competency development, and approved-tool guidance. Ethical Considerations appeared in 102 of 111 (91.9%) and three of six (50.0%; p =0.014), addressing algorithmic bias, misinformation, responsible use, accountability, intellectual property, and sustainability. Cognitive and Educational Considerations Cognitive and Educational Considerations appeared in 87 of 111 international sources (78.4%) and four of six Saudi sources (66.7%). Notably, explicit references to cognitive dependency or overreliance were identified in 47 international sources (42.3%) but in no Saudi source. Explicit treatment of overreliance appeared, for example, in Imperial College London’s Vision for Education in the Age of AI (2026) and Stanford University’s MD Program Policy 3.32. Privacy and Clinical-Data Governance Privacy and Data Protection appeared in 77 of 111 international sources (69.4%) and three of six Saudi sources (50.0%). Eight international universities included explicit clinical-data-governance provisions; Stanford University, Yale University, Johns Hopkins University, and Harvard Medical School specifically referenced protected health information (PHI) or HIPAA-regulated data, and several UK institutions referenced special-category health data under the GDPR. No Saudi governance source contained explicit provisions on clinical data governance, patient information, PHI, or healthcare-specific data protection. Health Informatics-Specific Governance Program-level governance specifically addressing AI use within Health Informatics curricula was identified at only two international institutions—Karolinska Institutet (Health Informatics Program AI Guidelines, 2025) and the University of California, San Francisco (Bridges Curriculum Generative AI Policy)—and at no Saudi university. Explicit alignment between AI governance and professional-competency or accreditation frameworks was identified at Stanford University, Karolinska Institutet, UCSF, and Imperial College London, and among no Saudi institutions. Thus, most institutions relied on university-wide governance rather than guidance tailored to Health Informatics. Discussion This study provides a comparative analysis of publicly available GenAI governance among global and Saudi universities offering Health Informatics and related programs. To our knowledge, it is among the first to systematically examine institutional governance of GenAI within Health Informatics education. Based on publicly available documentation, international and Saudi institutions differed substantially in policy availability, governance-domain coverage, content depth, and the incorporation of educational, ethical, and Health Informatics-specific considerations. All international universities had publicly available GenAI governance, whereas fewer than half of Saudi universities did—findings that should be read as differences in publicly available documentation rather than as conclusive evidence about internal governance. The breadth of governance among international universities reflects the rapid institutional response that followed the public release of ChatGPT, with most sources published between 2023 and 2026. This trajectory is consistent with higher-education literature reporting accelerated policy development around academic integrity, assessment security, transparency, and responsible AI use,3–5,23 and with UNESCO’s emphasis on formal governance structures for responsible GenAI integration.8,9 Governance among Saudi universities appeared more limited: six sources across sixteen eligible institutions, with ten having no publicly accessible documentation. Several factors may plausibly contribute to lower public availability, and we offer these as interpretation rather than measurement. Saudi higher education is characterized by relatively centralized, ministry-level governance, such that AI guidance may be issued or maintained at national or system levels rather than published on individual university sites; some documentation may reside on Arabic-language pages or authenticated intranets not captured by public searches; and national AI and digital-health initiatives under Vision 2030, including SDAIA and national e-learning bodies, are recent, so their translation into published university-level governance may still be maturing. Whatever the cause, the absence of publicly available guidance can create uncertainty for students and faculty regarding acceptable use, disclosure, and ethical responsibilities, which underscores the value of transparent publication. The largest divergences—Transparency and Disclosure, AI Literacy and Capacity Building, and Ethical Considerations—were statistically significant on Fisher’s exact testing despite the small Saudi sample, indicating robust rather than incidental differences. Many international universities appear to have moved beyond a purely compliance-oriented view of AI governance toward educational approaches that build students’ capacity to use AI responsibly, consistent with contemporary frameworks emphasizing transparency, accountability, and AI literacy.8,9,24 Disclosure-based approaches, prominent among international institutions, may offer a balanced alternative to prohibition by encouraging honesty while preserving legitimate educational use.3,22 Even where coverage was comparable—Acceptable and Prohibited Use, Governance and Accountability—the orientation differed: Saudi sources emphasized compliance, sanctions, and monitoring, whereas international sources more often incorporated human oversight, faculty responsibilities, and educational guidance. This suggests that governance maturity cannot be judged by domain coverage alone; the depth, purpose, and educational orientation of provisions matter, echoing calls for governance that supports responsible engagement rather than relying solely on enforcement.3,4,23 The cognitive findings are particularly noteworthy. Although overall coverage was similar between groups, explicit attention to cognitive dependency and overreliance appeared only among international institutions (42.3% of sources) and in no Saudi source. This aligns with emerging evidence that heavy reliance on GenAI may affect critical thinking, independent reasoning, and problem-solving,13,28 and with recent evaluations documenting inconsistent GenAI performance on health-professions assessments, which reinforce the need to cultivate critical appraisal rather than passive acceptance of AI outputs.32,33 In our own prior work, educators described a paradox of AI dependency in higher education, further motivating explicit attention to overreliance in institutional governance.17 The Health Informatics-specific findings represent a central contribution. Despite the discipline’s close ties to AI, program-level governance tailored to Health Informatics was identified at only two international institutions and no Saudi university, and clinical-data-governance provisions were absent from all Saudi sources. This matters because Health Informatics graduates will engage with AI differently from many peers—working as clinical data stewards, health-information managers, digital-health analysts, and informatics-enabled clinicians who interact with clinical decision support, predictive analytics, health information systems, and AI-enabled healthcare technologies.18 Governance frameworks that address only general academic use may be insufficient to prepare graduates for the ethical, legal, and professional responsibilities of AI use in care settings, where the protection of PHI is a core obligation and international guidance for health consistently emphasizes privacy, accountability, and patient protection across the AI lifecycle.27 Implications for Policy and Practice These findings suggest concrete, actor-specific actions. Universities that lack formal GenAI governance should establish institution-wide policies addressing transparency, accountability, ethical use, and disclosure, and should extend governance beyond academic integrity to include AI literacy, critical thinking, and responsible use; they should also publish these documents openly, since transparency is itself a governance indicator. Accreditation bodies for Health Informatics and health-professions programs should incorporate discipline-specific AI-governance and clinical-data-governance expectations into program standards, thereby signaling that AI competencies are a required educational outcome. National policymakers—particularly in the Saudi context, given Vision 2030 and SDAIA priorities—can support universities with model frameworks, Arabic-language guidance, and expectations for public availability. Across all actors, governance should be treated as a dynamic instrument requiring continuous review as AI technologies evolve. Strengths and Limitations This study is among the first comparative policy analyses focused specifically on GenAI governance in universities offering Health Informatics programs; it includes 46 institutions and 117 governance sources, applies a structured yet emergent analytical framework, incorporates a formal inter-coder reliability assessment, and supports comparisons with Fisher’s exact testing. Several limitations should be acknowledged. First, the analysis was limited to publicly available documentation and may not capture internal policies, departmental guidance, or implementation practices that were not publicly accessible; the absence of a public document does not establish the absence of internal governance, and reliance on public sources introduces potential publication bias. Second, the Saudi group comprised only six sources, so Saudi percentages are descriptive and sensitive to individual sources despite the supporting Fisher’s exact tests. Third, the composite maturity framework was developed for this study and has not been externally validated. Fourth, GenAI governance evolves rapidly; sources published after the May 2026 collection window were not captured, and findings represent a cross-sectional snapshot. Finally, the study examined the existence and content of documentation rather than its implementation or effectiveness. Future research should examine implementation and compliance, incorporate student and faculty perspectives, validate maturity instruments, and—through direct institutional surveys—capture internal governance not publicly available. Conclusion Based on publicly available documentation, international universities demonstrated broader and deeper GenAI governance than Saudi universities offering Health Informatics and related programs, with the largest and statistically significant differences in transparency, AI literacy, and ethical governance, and pronounced gaps in Health Informatics-specific and clinical-data governance. International institutions increasingly combine accountability with educational enablement, whereas publicly available Saudi governance remains more limited and compliance-oriented. These conclusions describe publicly available institutional documentation, which may not fully represent internal governance practices. Given the rapid evolution of GenAI, institutions should treat governance as requiring continuous review, and the governance elements identified here—transparency, AI literacy, ethics, and discipline-specific and clinical-data provisions—are broadly applicable internationally to universities seeking to support the responsible integration of GenAI in higher education and the preparation of future digital-health workforces.