The study, led by Joseph Conteh of the University of New Mexico together with Naim Kapucu, Eunice F. Ebevuhe, and Oluwapelumi Wumi Adeniji, was conducted and reported in accordance with the PRISMA 2020 reporting guideline, with study selection governed by predefined PICOS eligibility criteria. The team searched four databases—PubMed, Web of Science, Scopus, and Google Scholar—for peer-reviewed articles published from January 2017 through January 2026, with final searches completed between January 15 and 18, 2026. Of 1,065 records identified, just 22 studies met the eligibility criteria. Methodological quality was appraised with design-appropriate tools: the Critical Appraisal Skills Programme checklist for qualitative and certain observational designs, the Newcastle-Ottawa Scale for cohort and case-control studies, and QUADAS-2 for diagnostic accuracy studies. A narrative synthesis was followed by a structured regulatory gap analysis of federal statutes, regulations, and agency guidance, in which a gap was recorded only when at least two independent sources converged. Inter-rater agreement at the title and abstract screening stage was substantial, with Cohen’s kappa of 0.84.

The review’s central diagnosis is that the federal regulatory architecture was built for a fundamentally different technological paradigm. Authority over hospital AI is dispersed across agencies that each hold a partial slice of oversight: the FDA evaluates safety and effectiveness of Software as a Medical Device through pre-market review; the Office of the National Coordinator for Health Information Technology certifies health IT and mandates transparency for predictive decision-support interventions through the HTI-1 rule; the Centers for Medicare and Medicaid Services shapes adoption through conditions of participation, quality reporting, and payment policy; HIPAA governs the privacy and security of health data; and the Federal Trade Commission polices deceptive or unfair AI claims. In a decentralized federal environment, state legislation has emerged as a supplementary force, producing a patchwork landscape—California, for example, has established some of the nation’s strictest privacy and transparency standards, requiring healthcare institutions using AI-assisted diagnostic tools to inform patients about the system’s functionality, diagnostic basis, and potential risks.

The most technically striking finding concerns the chasm between pre-market validation and real-world performance. A COVID-19 chest radiograph AI diagnostic tool that achieved an area under the receiver operating characteristic curve greater than 0.95 on open datasets dropped to approximately 0.70 after prospective deployment across multiple hospital settings, varying from 0.66 to 0.74 across time quartiles over 19 post-deployment weeks. Predictive models delivering AUC values of 0.96 for readmission prediction and 0.83 from roughly 280,000 cases were tested only on retrospective data without external validation. The review is careful to distinguish deficiencies in the evidence submitted to the FDA from deficiencies in FDA requirements themselves—many validation gaps originate in the pre-submission research pipeline rather than in statutory standards, a distinction that matters because the remedies differ. Even stringent validation systems such as APPRAISE-AI, which rates methodology, results, and reproducibility, do not fully account for temporal and situational variation in clinical settings.

When regulatory-grade evidence does exist, the results can be dramatic. The KATE triage system outperformed nurses by 26.9% in general acuity assignment across 729 reviewed patient records, decreased undertriage from 19.8% to 9.7% and overtriage from 20.4% to 14.4%, and captured 1,880,841 features at 12.79 per encounter with 98.47% accuracy. An FDA-authorized AI tool for pulmonary embolism detection achieved 99.9% negative predictive value across 11,736 real-world imaging tests and slashed worklist detection times from 7,714 minutes to 87 minutes while reducing missed cases from 44.8% to 2.6%. A machine learning sepsis prediction algorithm enabled antibiotic delivery within 2.76 hours, cutting in-hospital mortality from 21.3% to 9.0%—a 58% relative reduction. But these successes remain exceptions rather than the norm, and the review found that AI’s relationship with CMS quality metrics is mixed: in a longitudinal observational study spanning 70 CMS quality indicators from 2022 to 2025, hospitals using AI showed positive trends for pneumonia mortality and hospital-acquired conditions but negative trends for readmission rates and sepsis treatment compliance. Because these are observational associations, they do not establish causal effects.

The structured gap analysis identified four critical, structurally interconnected failures. The first is the post-market monitoring gap: only 58.1% of hospitals in a national sample performed post-implementation monitoring in 2024, and the majority of those lacked continuous drift detection. In related survey data, 47.4% of hospitals did not evaluate post-deployment accuracy, 51.4% had no bias assessment protocols, and in 61% of hospitals post-deployment accuracy was assessed but bias was not, with all monitoring remaining sporadic. The second is the equity assurance gap. The training dataset of a readmission prediction model included less than 2% Hispanic and Asian participants while Caucasian patients made up more than 75%, with no subgroups evaluated for differential performance. The KATE system showed an 11.4 percentage point accuracy disparity between adults (78.5%) and children (67.1%) without subgroup analyses for race, ethnicity, or socioeconomic status. The review also distinguishes algorithmic bias from structural adoption disparities: larger hospitals had a 15.2% higher probability of adopting machine learning than smaller ones, health system-owned hospitals had a 26.8% higher probability than independently owned ones, and regions of high healthcare need in low-adoption areas had persistently lower adoption probabilities—including ratios of 0.75 for primary care Health Professional Shortage Areas and 0.74 for high Area Deprivation Index areas.

The third failure is the accountability clarification gap, which the review dissects into four frequently conflated dimensions: legal liability, clinical responsibility, institutional governance, and vendor obligations. Radiologists bore responsibility for diagnosing pulmonary embolism, yet responsibility allocation for 12 cases of missed diagnosis remained unclear. An AI-clinician hybrid framework for COVID-19 diagnosis did not determine responsibility distribution among developers, informaticians, and treating physicians, and a randomized clinical trial of sepsis prediction specified no clear distribution of responsibility in case of patient harm. The problem is compounded by market concentration: 79% of hospitals use predictive models created by electronic health record vendors, yet no frameworks exist for determining accountability between hospitals, developers, vendors, and clinicians. The fourth is the governance capacity gap, which operates as a cross-cutting constraint across the others. Only 65% of hospitals used predictive models and only 44% assessed biases, with higher implementation rates concentrated among larger and more profitable institutions. Even best-practice tools illustrate the barrier: DEPLOYR required Azure Functions, specialized Python libraries, data warehouse access with institutional review board approval, and registration at the Epic App Orchard, while the MedPerf federated benchmarking initiative demanded collaboration between more than 20 firms, 20 universities, and 9 hospitals across 13 countries, encountering regulatory, legal, and reputational barriers throughout.

The documented harms lend urgency to these technical findings. The widely cited study by Obermeyer and colleagues demonstrated that a commonly used healthcare algorithm systematically underestimated illness severity among Black patients, resulting in only 17.7% being identified as eligible for advanced care programs when a corrected, unbiased algorithm would have identified 46.5%. Insufficiently externally validated sepsis prediction technologies have generated alert fatigue that paradoxically diminishes clinician responsiveness to genuine patient deterioration signals. Survey data show that 73.1% of clinicians regarded lack of trust, transparency, or bias as critical barriers to implementation, and evaluations of large language models found that GPT-4 provided consistent responses in only 67 of 80 clinical scenarios (83.8%), with both GPT-3.5 and GPT-4 demonstrating weakness in recognizing diversity, equity, and inclusion principles. The review notes that the HTI-1 rule’s transparency mandates, while a positive step, suffer from underdeveloped enforcement mechanisms and unstandardized disclosure content.

International comparisons sharpen the picture. The European Union Artificial Intelligence Act imposes binding pre-market conformity assessment for high-risk AI systems, including most medical AI, and centralizes enforcement in designated market-surveillance authorities while assigning explicit obligations to both providers and deployers. U.S. requirements, by contrast, remain dispersed across FDA, ONC, and CMS authorities, with enforcement divided among federal and state actors holding partial jurisdiction, and post-market mechanisms such as Predetermined Change Control Plans still in early implementation. Drawing on these models, the authors propose a lifecycle governance framework with six evidence-linked components: rigorous pre-market clinical validation with standardized benchmarking, mandatory post-deployment monitoring with automated drift detection and performance alerts, standardized transparency and explainability aligned with the HTI-1 rule, equity-centered assessment protocols requiring subgroup analysis and bias reporting, clear accountability frameworks delineating responsibility among developers, vendors, institutions, and clinicians, and scalable governance infrastructure with targeted support for under-resourced settings.

The authors acknowledge key limitations: the retrospective registration of the review protocol on the Open Science Framework, the restriction to English-language publications, database coverage limited to four sources while Embase, CINAHL, and ClinicalTrials.gov were excluded, and the heterogeneity of the included evidence, much of it retrospective and single-center. International studies from Canada, Saudi Arabia, and the Netherlands were used to contextualize rather than as direct evidence of U.S. regulatory performance. Still, the core conclusion stands on firm ground: industry projections indicate AI-enabled technologies will influence decision-making in most major U.S. health systems within the coming decade, and if the identified gaps persist as deployment scales, the risks to patient safety, equity, and clinical trust will grow commensurately. The review closes by grounding its call in medicine’s oldest ethical principle—primum non nocere, first, do no harm—arguing that only a national, lifecycle-focused, systems-based governance approach can ensure that hospital AI remains safe, ethical, and trustworthy as it reshapes clinical care.

Subject of Research: A systematic review and regulatory gap analysis of AI governance in U.S. hospitals, identifying four critical regulatory gaps and proposing a lifecycle governance framework.

Article Title: A framework for safe, ethical, and trustworthy AI deployment and integration in hospitals: a systematic review and regulatory gap analysis with emphasis on the U.S. regulatory context

Article References: Conteh, J., Kapucu, N., Ebevuhe, E. F., & Adeniji, O. W. (2026). A framework for safe, ethical, and trustworthy AI deployment and integration in hospitals: a systematic review and regulatory gap analysis with emphasis on the U.S. regulatory context. Frontiers in Digital Health, 8, Article 1955958. https://doi.org/10.3389/fdgth.2026.1955958

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1955958

Keywords: AI governance, hospital AI, FDA regulation, health equity, patient safety, systematic review, regulatory gaps, digital health, framework, safe, ethical, trustworthy

The study, conducted by Fei Song and Michael Dunn of the Centre for Biomedical Ethics at the National University of Singapore together with Zhenni Hu of Nanjing Normal University, examined 43 articles indexed in the Chinese Social Sciences Citation Index along with four policy reports. The analysis reveals that Chinese scholars have made substantial contributions to policy and governance discussions around trustworthy AI, but that these contributions remain conceptually scattered, with different disciplines approaching the question of trust from radically different angles. The researchers set out to synthesize this dispersed knowledge into a coherent conceptual account, and in doing so they arrived at a framework that treats trustworthiness not as a property of algorithms alone but as an emergent condition spanning technology, institutions, and human relationships.

The first layer of the framework concerns technical foundations. This is the layer most familiar to engineers and computer scientists, encompassing properties such as reliability, robustness, transparency, explainability, fairness, and safety. Chinese-language scholarship in this domain is extensive and spans a striking range of applications. Researchers have examined explainable learner models as the technical key to trustworthy personalized learning, explored blockchain-driven approaches to financial security intelligence, and investigated how algorithmic transparency in personalized recommendation systems shapes users’ perceptions of trustworthiness. Work on data annotation governance has highlighted backstage risks in the labor-intensive processes that feed machine learning systems, while archival science perspectives have been brought to bear on the question of algorithm provenance, asking how the origins and histories of algorithms can be documented and verified. These contributions demonstrate that the technical layer of trustworthiness is not merely a matter of model architecture but extends deep into the data pipelines, documentation practices, and supply chains that underpin AI systems.

The second layer addresses governance and regulatory arrangements. Here the Chinese literature is particularly distinctive, reflecting the country’s active policy discourse on AI regulation. Scholars have assessed pathways for trustworthy AI legislation, explored innovation-friendly regulatory designs that balance safety with technological development, and analyzed governance practices in terms of models, actors, objects, and tools. Policy documents play an important role in this layer, including the Ministry of Science and Technology’s 2019 governance principles for developing responsible AI, a white paper on trustworthy artificial intelligence produced by the China Academy of Information and Communications Technology, and more recent industry ecosystem reports and governance surveys. The review shows that Chinese scholars have been grappling with the same fundamental questions that animate regulators in Europe and the United States: how to translate abstract ethical principles into enforceable rules, how to assign liability when AI systems fail, and how to design oversight mechanisms that keep pace with rapidly evolving technology. One striking example is legal scholarship on liability when human–machine co-piloting fails, examining the negligence of safety operators through the lens of the principle of reliance.

The third layer, and the one the authors argue is most often neglected in purely technical accounts, concerns the conditions of human–AI interaction under which trust is formed and calibrated. The Chinese-language literature here is remarkably rich, drawing on empirical psychology, communication studies, and philosophy. Researchers have studied trust in automated vehicles, safety trust in intelligent domestic robots, and human–AI mutual trust in the era of artificial general intelligence. A dual-pathway model of trust calibration distinguishes between trust dampening and trust promoting mechanisms, offering a nuanced picture of how users adjust their reliance on machines. Other studies examine how anthropomorphism affects trust in AI agents, with a meta-analysis synthesizing evidence on how humanlike features in AI systems influence trust and how contextual factors moderate that effect. Work on large language models has explored how machine hallucinations impair human–machine trust and what recalibration mechanisms might restore it, a question of obvious urgency as chatbots and AI assistants become ubiquitous.

What emerges from this synthesis is the insight that these three layers are not independent. An AI system can be technically excellent, with high accuracy and robust performance, and still fail to be trustworthy if the governance arrangements around it are inadequate or if the conditions of human interaction systematically miscalibrate trust. Conversely, strong institutions and well-designed interaction conditions cannot compensate for fundamentally unreliable technology. The authors’ central claim is that an AI system is trustworthy only if it meets the requirements at all three layers simultaneously. This conjunctive structure has practical implications: it means that trustworthiness cannot be certified by any single test, audit, or label, but requires integrated assessment across the full stack from model internals to regulatory environments to the design of the encounters between humans and machines.

The framework also speaks to a long-running philosophical debate about whether trust in AI is even coherent. Some Western philosophers argue that trust, properly understood, requires mutual expectations and moral agency that machines cannot possess, and that what we call trust in AI is really a form of reliance or confidence. Others contend that it is entirely possible to trust medical AI systems and that denying this misunderstands the functional role of trust in clinical decision-making. The Chinese literature adds distinctive voices to this debate, with scholars asking directly whether artificial intelligence can serve as a trustee and defending the notion of trustworthy AI from a logical perspective. The review also engages with a network account of trustworthy AI, which holds that trustworthiness is distributed across the networks of actors, artifacts, and institutions surrounding an AI system rather than residing in the system alone. The three-layer framework can be read as a structured elaboration of this network idea, specifying the technical, institutional, and interactional nodes that such a network must contain.

The methodological significance of the study lies partly in its source base. Scholarship published in Chinese remains largely absent from mainstream academic debates, a gap reinforced by publication incentives that push Chinese humanities and social science researchers toward international English-language venues. By drawing on the Chinese Social Sciences Citation Index, the authors accessed a body of work that includes empirical studies using artificial neural networks and Q methodology to map the dimensions of human–AI trust, philosophical analyses grounded in Confucian traditions, and domain-specific investigations in education, journalism, healthcare, and finance. The result is a more globally representative picture of what the research community actually knows about trustworthy AI. The authors also note distinctive features of the Chinese context, including a practice-oriented view of science and technology that shapes how trustworthiness is framed, and a governance culture in which state-led principles, industry white papers, and academic analysis interact closely.

The study concludes by identifying three directions for future research. First, the conceptual fragmentation of the field calls for continued theoretical work to integrate technical, governance, and interactional perspectives into unified accounts of trustworthiness. Second, the rapid emergence of generative AI and large language models creates new empirical challenges, from hallucination-induced trust erosion to the trust dynamics of human–machine dialogue, that demand updated frameworks. Third, cross-linguistic and cross-cultural comparative research is needed to test whether the three-layer structure holds across different scholarly traditions and regulatory cultures, or whether trustworthy AI means genuinely different things in different contexts. As AI systems increasingly mediate news consumption, medical diagnosis, education, and financial services, the stakes of getting trustworthiness right could hardly be higher. This review makes the case that answering the question well requires listening to conversations that have been happening in Chinese all along, and that the resulting framework, grounded in arguments advanced in the Chinese context, offers a more robust conceptual account of how disparate claims about trust fit together than either technical or governance approaches alone can provide.

Subject of Research: Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship

Article Title: Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship

Article References: Song, F., Hu, Z., & Dunn, M. (2026). Trustworthy AI beyond the technical: a three-layer framework informed by Chinese-language scholarship. AI & SOCIETY. https://doi.org/10.1007/s00146-026-03366-2

Image Credits: AI Generated

DOI: 10.1007/s00146-026-03366-2

Keywords: Trustworthy, beyond, technical, three-layer, framework, informed, Chinese-language, scholarship, scientific research