Patient, Caregiver and Frontline Participation in Clinical AI Governance in Viet Nam: From Consultation and Co-Design to Governance Input and Continuous Learning
IMPLEMENTATION INSIGHTS
Patient, Caregiver and Frontline Participation in Clinical AI Governance in Viet Nam: From Consultation and Co-Design to Governance Input and Continuous Learning
IMPLEMENTATION INSIGHT
Last reviewed: 5 October 2026
Healthcare organizations often discuss AI governance in terms of committees, policies, risk registers, validation reports and monitoring dashboards. Those structures matter, but they can still miss something fundamental: the people who experience the effects of AI in real care.
A clinical AI system may appear safe from a technical perspective while creating an impractical workflow for nurses. A generative tool may reduce documentation time while producing information that patients find confusing or difficult to correct. An alert may perform well overall but become disruptive in a particular specialty. A patient-facing AI pathway may be technically accessible yet difficult for older people, caregivers, people with limited digital literacy or those with particular communication needs. These effects may not appear in a conventional model-performance report, but they can determine whether AI ultimately improves care.
This is why participation should be treated as part of responsible AI governance rather than as an optional engagement or public-relations exercise. In September 2026, WHO/Europe published findings from its Knowledge Community on responsible AI in health. Participants identified participation by patients, communities and frontline professionals as one of the priority areas for responsible AI and described codesign with patients and clinicians as an important enabling condition [7]. The report reflects the perspectives of Knowledge Community participants and should not be interpreted as a formal normative position of WHO, but the underlying question is highly relevant to healthcare organizations in Viet Nam: have the people affected by an AI system had an appropriate opportunity to influence how it is selected, implemented, monitored and improved?
Participation is a continuum, not a single activity
The word participation can describe very different levels of involvement. Informing a patient that AI is being used is not the same as asking for feedback. Asking staff to complete a survey is not the same as involving them in workflow design. Having one patient representative attend a meeting is not the same as ensuring that patient experience can meaningfully influence a governance decision.
For clinical AI, a useful participation continuum is inform → consult → involve → co-design → provide governance input → contribute to continuous learning. Different systems will justify different levels of participation. A low-impact administrative tool with no material patient-facing consequences may require relatively limited engagement, while a patient-facing system, a system influencing important clinical decisions or a technology that substantially changes how care is delivered may justify deeper involvement.
The objective is not to require every patient, caregiver or clinician to become an AI governance expert. It is to ensure that governance decisions are informed by the perspectives needed to understand how the technology actually affects care.
Patients, caregivers and frontline professionals contribute different kinds of knowledge
Patients, caregivers and frontline professionals should not be treated as interchangeable stakeholders. Patients experience AI through care. They may notice whether information is understandable, whether access becomes easier or harder, whether communication feels less personal, whether an AI-generated summary accurately reflects what was discussed or whether a new digital pathway creates additional burden at a moment when they are already vulnerable.
Caregivers may see a different part of the system. They may coordinate appointments, medicines, transport, home monitoring or communication across multiple providers. For children, older people, people with disabilities, patients with cognitive impairment or people requiring substantial family support, caregivers may identify practical risks that neither a technology team nor an individual clinical encounter reveals. Caregiver participation should, however, respect the patient's autonomy, decision-making capacity, confidentiality and applicable arrangements for representation. A caregiver's perspective should not automatically be treated as identical to the patient's own voice.
Frontline professionals contribute knowledge of how work is actually performed rather than only how a process is expected to work. Nurses, physicians, pharmacists, allied health professionals, technicians, care coordinators, quality staff and administrative teams may recognize that an AI output arrives too late to matter, an alert competes with more urgent tasks, a new workflow creates duplicate documentation or staff are developing informal workarounds simply to keep the service functioning.
CHAI's September 2026 discussion of its AI Governance Playbooks provides a useful practice example. A nursing contributor emphasized the importance of explicitly representing frontline clinical perspectives because different professions identify different effects on patient care, workflow and decision-making. The discussion describes involvement across evaluation, implementation, ongoing monitoring and decisions about whether a technology should continue to be used, rather than participation ending after procurement [11]. Participation matters precisely because different people often see different parts of the problem.
Viet Nam's AI law creates a human-centered context, but it does not prescribe a participatory governance model
Viet Nam's Law No. 134/2025/QH15 on Artificial Intelligence has been effective since 1 March 2026. Its fundamental principles include a human-centered approach, protection of human rights and privacy, and the principle that AI serves people rather than replacing human authority and responsibility. The Law requires human oversight and the capacity for human intervention in decisions and actions of AI systems. For sectors with direct effects on life, health, lawful rights and interests or social order and safety, AI is subject to stricter risk management appropriate to the characteristics of the sector. In health care specifically, the Law refers to patient safety, reliability under real-world conditions of use and protection of health data [1].
The Law also defines an “affected person” as an organization or individual whose lawful rights and interests, life, health, property, reputation or access to services are directly or indirectly affected by the deployment or output of an AI system [1]. For high-risk AI systems, providers and deployers have defined transparency and accountability responsibilities. Among these, they must provide users and affected persons with publicly available information at a functional description level about system operation and risk warnings to support safe use [1].
Decree No. 142/2026/NĐ-CP, effective since 1 May 2026, further reflects the relevance of affected populations. Risk-classification documentation for medium- and high-risk AI systems includes information on the system's deployment context, including users and affected persons. The Decree also provides that transparency responsibilities should be implemented in a manner appropriate to the system's purpose of use, deployment context and level of risk [2].
These provisions are important, but they should not be overstated. As of this review, Vietnamese AI law does not establish a universal requirement for healthcare organizations to place patients, caregivers or frontline representatives on AI governance committees, conduct co-design for every clinical AI system or establish a particular patient-participation structure for AI governance. The participation approaches discussed in this article should therefore be understood primarily as responsible healthcare governance, implementation and patient-safety practices that can complement legal compliance, rather than as an additional statutory checklist.
Participation is not the same as consent
Participation and consent are related but fundamentally different concepts. A patient may contribute to designing or evaluating an AI-enabled care pathway without thereby providing legal consent for every use of personal data within that pathway. Conversely, obtaining legally valid consent for personal-data processing does not mean that the healthcare organization has meaningfully involved patients in decisions about how an AI system should be designed or used.
Viet Nam's Law No. 91/2025/QH15 on Personal Data Protection and Decree No. 356/2025/NĐ-CP have both been effective since 1 January 2026 [3,4]. Where consent is required under the applicable personal-data framework, the Law requires consent to be clear, specific and capable of being recorded or verified, and silence or lack of response does not constitute consent. Decree No. 356 further provides that organizations responsible for obtaining consent must retain evidence of it, prohibits default-consent mechanisms or unclear instructions that could cause confusion between consent and non-consent, and requires data subjects to be informed when consent is sought for the processing of sensitive personal data [3,4].
Participation, transparency, consent and the legal conditions governing personal-data processing are therefore separate questions that should be considered explicitly rather than treated as substitutes for one another. The participation process itself may also generate personal data through interviews, surveys, complaints, recordings or other forms of feedback, and those data remain subject to applicable protection requirements.
Participation should begin before procurement
One of the most valuable moments to involve people is before an AI product has been selected. Once a contract has been signed and a deployment timetable announced, participation can easily become an exercise in persuading staff or patients to accept a predetermined solution. Earlier involvement allows a more fundamental question: what problem are we actually trying to solve?
Leadership may believe that a clinical department needs an AI prediction tool, while frontline staff reveal that the main problem is delayed access to laboratory results. A hospital may plan an AI chatbot to improve navigation while patients explain that the real difficulty is uncertainty about where to go after a referral. A vendor may offer an automated documentation system while clinicians identify a broader problem in how information moves between specialties.
Participation before procurement can therefore reduce the risk of using AI to optimize the wrong problem. Patient and caregiver input will not be necessary for every procurement decision, but where AI materially changes patient communication, clinical decision-making, access to care, surveillance, documentation or the patient journey, organizations should consider whether the people affected by the technology understand the problem differently from those purchasing it.
Frontline involvement is particularly important because staff can identify practical requirements that may never appear in a conventional technical specification: when an output must arrive, who needs to see it, which systems it must integrate with, what competing alerts already exist, what information clinicians need in order to verify a recommendation and what should happen when the technology is unavailable.
Co-design should focus on care, not just the interface
Co-design is sometimes reduced to asking users whether a screen is easy to navigate. Usability matters, but clinical AI changes more than interfaces. A useful co-design process asks how the technology changes the care pathway: who encounters the output first, whether responsibilities shift, whether the timing of a clinical decision changes, what additional information a clinician needs before acting, how a patient can ask for clarification, what happens when a clinician disagrees with the AI and whether new work is created for another department that was not involved in the original project.
The focus should therefore move from “Do users like the AI?” to “What does this technology change in the care system?”
WHO/Europe's April 2026 report, Artificial intelligence is reshaping health systems: state of readiness across the European Union, found that 81% of EU Member States were actively involving stakeholders in shaping AI governance and encouraged broader consultation with patients and the public. The report argued that meaningful public involvement can strengthen trust, support more equitable outcomes and help AI tools better reflect the needs of the people they are intended to serve [8]. These are European findings rather than Vietnamese legal requirements, but they reinforce the value of looking beyond technical deployment.
In Viet Nam, co-design should be proportionate to organizational reality. A large tertiary hospital may be able to conduct multidisciplinary design workshops or establish patient advisory mechanisms, while a smaller specialty organization may use structured interviews, workflow observation and targeted user testing. The principle matters more than the format: people affected by a material change in care should have a realistic opportunity to identify its consequences before those consequences become routine practice.
Frontline participation is particularly important during local validation
Technical validation can help determine whether an AI system produces sufficiently reliable outputs under defined conditions. Frontline participation can reveal whether those outputs can actually be used safely in practice.
An alert may be accurate but arrive at a time when nobody can respond. An AI-assisted imaging prioritization system may work technically but disrupt an existing escalation pathway. A generative clinical note may appear fluent while requiring so many corrections that documentation burden increases. A system may perform well during a closely supervised pilot but be used differently during night shifts, weekends or periods of high workload.
Local validation should therefore consider not only model performance, but also human-AI performance within the real workflow. Frontline professionals can help identify what needs to be observed: response time, overrides, ignored alerts, workarounds, duplicate tasks, clarification requests, additional handoffs, workload changes and effects on existing safety checks. They may also recognize when behavior during a pilot differs from normal practice because users know they are being closely observed.
This does not mean replacing quantitative validation with user opinion. It means combining technical, clinical, operational and experiential evidence to understand how the system performs in real care.
Governance input does not mean that everyone needs formal decision authority
Meaningful participation should not create confusion about accountability. A healthcare organization may appropriately seek input from patients, caregivers or frontline professionals without giving every participant formal authority to approve, reject or suspend an AI system. Different governance structures can allocate advisory, consultative, technical, clinical, executive and decision-making roles differently.
The important question is whether input has a realistic route to influence a decision. A multidisciplinary AI governance group might receive structured feedback from patients when evaluating a patient-facing system. Nurses, physicians or other users from the affected service may participate directly in evaluation meetings. Quality and patient-safety teams may have authority to escalate safety concerns, while final organizational approval remains with an appropriately authorized leader or governance body.
What should be avoided is participation without influence: consulting people after a decision has effectively become irreversible, repeatedly collecting feedback without explaining what happened to it, or inviting a representative into discussions without giving that person understandable information or a realistic mechanism for raising concerns.
A simple governance discipline can help: What did we hear? What did we decide? What changed because of the input? If nothing changed, why not? Documenting that sequence turns participation into governance evidence rather than symbolism.
Patients and caregivers may identify risks that conventional monitoring misses
Clinical AI monitoring often focuses on technical performance, drift, incidents and system availability. Patients and caregivers may observe different signals.
An automated communication may sound authoritative even when the underlying information is uncertain. A patient may believe an AI-generated message came directly from a physician. An ambient documentation system may change the character of a consultation because a patient is uncomfortable with audio capture or recording. A portal may introduce an AI assistant that works well for digitally confident patients but makes navigation more difficult for others.
Caregivers may notice that AI-generated advice does not fit what happens at home, that different parts of the care pathway provide conflicting instructions or that automation increases rather than reduces the coordination burden placed on families. These observations should not automatically be interpreted as evidence that an AI system is unsafe; they should become inputs to evaluation.
Patient experience, complaints, requests for record correction, call-center patterns, opt-out behavior where applicable, accessibility concerns and qualitative feedback can all help an organization understand effects that conventional model metrics may not detect.
Joint Commission's 2026 Responsible Use of AI in Healthcare certification provides a relevant international organizational reference point. The voluntary U.S. certification recognizes organizations that demonstrate governance, safeguards, monitoring processes and education for responsible AI use. Its standards include support for patient transparency and/or patient consent where appropriate and relevant education for patients and families, while treating responsible AI as an issue of patient safety, quality, governance, privacy and trust rather than technology alone [9]. It is not a Vietnamese legal standard, but its organizational framing is useful.
Representation matters because “the patient voice” is not one voice
Participation can unintentionally reinforce inequity when organizations hear mainly from the people who are easiest to engage. Patients with strong digital literacy, good health knowledge and enough time may be more likely to participate in advisory groups. Yet some of the people most affected by poorly designed AI may be harder to reach, including older people, people with disabilities, people with limited literacy, people who do not routinely use digital services, individuals facing communication or language barriers, and patients whose care depends substantially on family support.
Viet Nam's AI Law adopts a human-centered approach and specifically prohibits developing, providing, deploying or using AI systems for the purpose of exploiting vulnerabilities of groups including children, older people, people with disabilities, ethnic minorities, people who lack or have limited civil act capacity and people with cognitive or behavioral difficulties in ways that cause harm to themselves or others [1]. This does not create a requirement for every AI project to include representatives of every group, but it reinforces a broader governance principle: participation should consider who is actually exposed to the system and who may experience disproportionate consequences if it fails.
Organizations should therefore ask whether the people consulted reasonably reflect the people affected. If a technology is primarily intended for older patients, user testing conducted almost entirely with younger healthcare staff is not adequate evidence of patient usability. If a system is intended for multiple sites with different populations and workflows, participation from a single flagship facility may not reveal what will happen elsewhere.
Frontline participation should include more than physicians
Clinical AI governance can easily become physician- and technology-centered. Physicians are essential participants, but AI often changes the work of nurses, pharmacists, technicians, administrative teams, care coordinators, allied health professionals and quality personnel as well.
An algorithm may alter clinical decision-making for a physician while simultaneously increasing documentation work for a nurse, changing medication verification for pharmacy or generating new escalation tasks for an administrative team. CHAI's September 2026 nursing perspective is particularly useful here because it emphasizes explicitly identifying relevant professions rather than assuming that their perspectives will automatically be represented. It describes AI governance as multidisciplinary and discusses mechanisms such as focus groups and advisory councils through which end users can evaluate and provide feedback on AI-enabled technologies throughout their lifecycle [11].
For healthcare organizations in Viet Nam, the practical principle is straightforward: map participation to the workflow, not to organizational hierarchy. Ask who touches the process before AI, who will touch it afterward and who will manage the consequences when the system behaves unexpectedly. Those are the perspectives that need a route into governance.
Participation should continue after go-live
Some of the most important information about an AI system appears only after routine use begins. Staff become more familiar with the technology, initial vigilance may decline, workarounds emerge, new users join, patient populations change, vendors release updates and workflow pressures vary across shifts and sites. Over time, people also learn what the system does well and where it is consistently weak.
Participation should therefore not end with a launch survey. A mature organization creates recurring ways for frontline teams to report what is changing and for patient or caregiver experience to reach the appropriate governance function. These mechanisms may include structured post-implementation reviews, recurring user discussions, integration with quality meetings, complaint analysis, patient advisory mechanisms, incident review or targeted evaluation after significant updates.
The feedback route should remain proportionate. Not every observation needs escalation to an executive AI governance body, and many issues can be resolved locally. What matters is that meaningful signals have a route upward when local teams cannot resolve them, when patient safety may be affected or when a pattern could be relevant to other departments or sites.
Frontline and patient feedback should connect with existing quality and safety systems
Healthcare organizations should avoid creating an isolated AI-feedback ecosystem. Circular No. 19/2013/TT-BYT remains recorded as in force and establishes a framework for hospital quality management that includes patient-centeredness, continuous quality improvement and participation of hospital staff in quality-management activities [5]. These existing structures provide an important foundation for integrating AI-related learning into routine quality work.
Circular No. 43/2018/TT-BYT establishes a framework for reporting, analyzing, responding to and preventing medical incidents and expressly recognizes near misses. The national legal database currently records the Circular as partially expired, so individual provisions should be applied according to their current legal status and together with later applicable law [6].
If a nurse reports that an AI warning repeatedly arrives only after clinical deterioration is already evident, that concern should not disappear into an IT ticket. If patients repeatedly request corrections to AI-generated documentation, the pattern should not remain isolated within customer service. If clinicians repeatedly override a system for the same reason, that pattern may represent a quality or safety signal rather than simple user resistance.
Participation becomes most valuable when these observations can enter existing mechanisms for quality review, patient-safety investigation, corrective action and remeasurement.
Psychological safety matters for frontline participation
An organization can establish a feedback channel and still receive very little useful information if staff believe that questioning an AI system will be interpreted as resistance to innovation. Frontline professionals need to be able to say that a system is inconvenient, confusing or potentially unsafe without automatically being labelled “anti-technology.” At the same time, governance should distinguish genuine safety concerns from implementation problems, lack of familiarity, insufficient training or ordinary resistance to change.
This requires a culture in which disagreement can be investigated rather than dismissed. The reverse also matters: staff who identify meaningful benefits should be heard. Participation is not designed to collect objections until an AI system is rejected; it is intended to improve organizational understanding.
A more useful governance question is therefore not simply “Do staff support the AI?” but “What are staff observing, and what evidence would help us understand whether those observations represent a safety, workflow, training or adoption issue?”
Participation should not become symbolic
Poorly designed participation can consume time without improving governance. Common failure modes include inviting stakeholders only after procurement, asking questions so broad that the answers cannot influence a decision, relying repeatedly on the same representative, using consultation primarily to legitimize an already approved project, requesting feedback but never explaining what happened to it, or asking people to participate in technical discussions without providing information in a form they can meaningfully use.
Another failure mode is confusing volume with representation. A thousand survey responses do not necessarily provide better governance insight than a carefully designed discussion with the people most affected by a specific workflow.
Participation quality should therefore be judged by practical questions: Was the right group involved? Was involvement early enough to matter? Did participants receive understandable information? Was the question relevant to their experience? Was there a route from input to decision? Did the organization explain what changed? If those conditions are absent, the organization may be collecting opinions without materially improving governance.
A practical participation architecture for clinical AI
Healthcare organizations in Viet Nam do not need a complex participation bureaucracy for every AI system. A simpler, risk-based architecture can begin by identifying who is affected: patients, caregivers, clinicians, nurses, pharmacists, technicians, administrators, quality teams or other groups whose work, decisions, data or care experience may materially change. The organization should then define why participation is needed. Participants may be helping to define the original problem, assess usability, evaluate workflow, identify risks, review patient communication, interpret post-deployment signals or contribute to a scale-up decision.
The next question is the appropriate level of involvement. Some situations may require information only; others may justify consultation, structured involvement, co-design or recurring governance input. Participation also needs to be accessible. Patients should not need expertise in machine learning to explain why an AI-enabled pathway is confusing or burdensome, and frontline professionals should not need to understand model architecture to describe how a system changes clinical work.
Important input should then be linked to a decision, action or documented rationale for taking no action. Feedback should connect with monitoring, quality and patient-safety systems so that it becomes one source of real-world evidence alongside performance data, incidents, complaints, audits, outcomes and workflow measures. Participation may also need to be repeated when circumstances materially change, such as after major system updates, deployment to new sites, expansion to different patient groups or substantial workflow changes.
The practical sequence can therefore be summarized as identify → purpose → involve → understand → decide → act → learn → re-engage. This participation architecture is proposed here as a practical governance approach; it is not a statutory framework under Vietnamese law.
Participation should be proportionate to risk and impact
A community clinic using a limited administrative AI function does not need the same participation infrastructure as a hospital group introducing AI across diagnostic, documentation and patient-facing systems. The depth of participation should generally increase where AI has greater clinical consequences, affects larger populations, substantially changes workflow, involves vulnerable groups, communicates directly with patients, uses sensitive information, creates difficult-to-reverse effects or introduces meaningful uncertainty.
Organizations should also avoid assuming that meaningful participation always requires a formal committee. Structured interviews, workflow observation, simulation, focus groups, patient-experience reviews, targeted usability testing and post-implementation discussions can all provide useful governance input when used appropriately. The aim is not to maximize the number of meetings, but to obtain the right information from the right people at the right time.
Participation and accountability must remain distinct
One boundary should remain clear throughout this process: patients, caregivers and frontline staff can contribute valuable governance knowledge, but participation does not transfer institutional responsibility to them.
A healthcare organization cannot justify an unsafe deployment by saying that clinicians were consulted. Support from a patient advisory group does not replace validation. A frontline vote cannot eliminate the need for legal compliance or clinical-risk assessment. A patient who agrees to use an AI-enabled service does not thereby assume responsibility for system safety.
Viet Nam's AI Law expressly provides that AI serves humans rather than replacing human authority and responsibility and requires the maintenance of human oversight and the capacity for human intervention [1]. Participation should therefore strengthen accountable decision-making, not distribute accountability so broadly that nobody remains responsible.
What this means for healthcare organizations in Viet Nam
Clinical AI governance is often described as something experts do: legal teams interpret regulations, technical teams evaluate systems, quality teams monitor performance and executives make decisions. All of those roles matter, but responsible governance also requires an organization to understand what AI looks like at the point where it actually enters care.
What does the patient experience? What burden falls on the caregiver? What does the nurse see during a busy shift? What happens when AI conflicts with a clinician's judgment? Which workaround has quietly become routine? Which group finds the new pathway harder rather than easier? What signal becomes visible to frontline staff before it appears on a monitoring dashboard?
These are not secondary questions. They are part of determining whether technology is functioning safely and effectively within the health system. The strongest governance model is therefore not one in which every stakeholder makes every decision. It is one in which relevant voices can reach the decisions that affect them, evidence and accountability remain clear, and real-world experience feeds back into continuous improvement.
For healthcare organizations in Viet Nam, the transition can be expressed simply: from informing people about AI → to listening to people affected by AI → to involving them where their experience can improve governance → to learning continuously from what happens in real care. That is where participation becomes more than consultation and becomes part of responsible healthcare AI.
Professional education notice
This article is intended for professional education and knowledge exchange. It does not constitute individualized legal, regulatory, clinical, research-ethics, data-protection, accreditation or quality-management advice.
The participation approaches described in this article, including consultation, involvement, co-design, governance input and continuous learning, are practical healthcare governance and implementation concepts. They are not presented as universal statutory requirements under Vietnamese law.
Nothing in this article should be interpreted as creating a general legal requirement for patients, caregivers or frontline professionals to hold formal membership or decision-making authority on healthcare AI governance committees. The appropriate form of participation depends on the AI system, clinical context, affected population, organizational structure and applicable law.
Patient or caregiver participation in AI governance should also not be confused with consent for personal-data processing, consent to clinical care, legal representation or authorization to act on behalf of another person. Those questions should be assessed separately under the applicable legal and institutional framework.
References
National Assembly of Viet Nam. Law No. 134/2025/QH15 on Artificial Intelligence. Adopted 10 December 2025; effective 1 March 2026.
Government of Viet Nam. Decree No. 142/2026/NĐ-CP Detailing a Number of Articles and Measures for Implementation of the Law on Artificial Intelligence. Issued 30 April 2026; effective 1 May 2026.
National Assembly of Viet Nam. Law No. 91/2025/QH15 on Personal Data Protection. Adopted 26 June 2025; effective 1 January 2026.
Government of Viet Nam. Decree No. 356/2025/NĐ-CP Detailing a Number of Articles and Measures for Implementation of the Law on Personal Data Protection. Issued 31 December 2025; effective 1 January 2026.
Ministry of Health of Viet Nam. Circular No. 19/2013/TT-BYT Providing Guidance on Management of Quality of Medical Examination and Treatment Services in Hospitals. Issued 12 July 2013; effective 15 September 2013; currently recorded as in force.
Ministry of Health of Viet Nam. Circular No. 43/2018/TT-BYT Providing Guidance on Prevention of Medical Incidents in Medical Examination and Treatment Establishments. Issued 26 December 2018; effective 1 March 2019; currently recorded in the national legal database as partially expired.
World Health Organization Regional Office for Europe. Report of the Knowledge Community on Responsible Artificial Intelligence in Health. Published 1 September 2026.
World Health Organization Regional Office for Europe. Artificial Intelligence Is Reshaping Health Systems: State of Readiness across the European Union. Published 20 April 2026.
Joint Commission. Responsible Use of AI in Healthcare Certification. Launched 1 June 2026.
Coalition for Health AI. AI Governance Playbooks. Released 27 May 2026.
Coalition for Health AI. Governance Playbook Spotlight: A Nursing Perspective on Putting AI Governance Into Practice. Published 9 September 2026.