.png)
AI ETHICS TOOLKIT
PIONEERING ARTIFICIAL INTELLIGENCE IN NURSING
There are many complex ethical issues related to using artificial intelligence based models and tools in healthcare such as privacy and trust, transparency, accountability and more. The AI Nurses Network will be launching the world's first AI ethics toolkit for nurses here very soon!
AI Ethics Toolkit
Understanding the complexities of AI in nursing and midwifery
We have recently completed a large systematic review (220 studies) examining the ethical dimensions of artificial intelligence (AI) in nursing and midwifery practice (i.e., education, clinical practice/patient care and research). The pre-print manuscript is available online here.
We found 6 key themes and 20 subthemes outlined in the practical toolkit below. This will help nurses understand the various ethical issues with using AI systems and tools, and some of the governance and other strategies that are needed to work with AI in professional practice.
Theme 1: Preserving humanity
Care ethics in AI-mediated nursing and midwifery practice
1.1 Relational care under threat
AI systems that prioritise measurable patient and health service outcomes without also measuring direct care and its outcomes may lead to hands-on care and therapeutic relationships becoming less valued and less visible. These efficiency-driven models of care could negatively affect the experience of caring and reduce holistic and person-centred care (Issa et al., 2024 ; Tuncer & Tuncer, 2024).
1.2 Value alignment as ethical imperative
AI models and tools such as generative AI should be grounded in core nursing and midwifery values (e.g., empathy, compassion, dignity) to preserve the moral foundations and practices of the caring professions (Ramadan et al., 2024 ; Watson, 2024).
1.3 The authenticity dilemma
Uncertainty exists about whether AI-generated empathy and other emotions constitutes caring and whether vulnerable populations may misinterpret artificial empathy as real, potentially affecting autonomy and trust (Metzler & Barnes, 2014 ; Montemayor et al., 2022).
Theme 2: Beyond technical reliability
Safety as an ethical and relational imperative
2.1 Premature deployment risks
Concerns around the safety and reliability of AI tools in clinical care have been raised, as AI can introduce errors and inaccuracies (including biases) in its outputs. Therefore, nurses and midwives remain accountable for AI-supported decision making and care delivery despite the professions lack of autonomy, authority or control over AI model or system development and performance (Davis, 2025; Issa et al., 2024).
2.2 Misinformation as patient harm
AI-generated misinformation is an ongoing safety concern including risks of inappropriate clinical decisions and care delivery, particularly if AI "hallucinations" create false but convincing information that may appear credible and authoritative despite being inaccurate (Luo et al., 2025). Nurses and midwives also reported challenges verifying the accuracy of AI outputs and the risk to patient safety (Coşkun et al., 2024), with the clinical expertise of both professions needed to protect patients, families and staff from AI-related informational harms.
2.3 Risk mitigation and navigation
Proactive risk mitigation strategies are needed for safe AI implementation such as robust cybersecurity and governance processes. Placing trust in nurses and midwives professional judgement rather than relying on the technical performance of AI models and tools alone is critical as well as preventing excessive alerts from AI systems which may reduce trust and hinder adoption of AI among the professions (Hassan & El-Ashry, 2024; Nashwan et al., 2023).
Theme 3: Autonomy, authority and accountability
Governance challenges and solutions
3.1 Patient autonomy, rights and dignity
AI use could threaten patient autonomy and confidentiality if their consent is not gained for its use in clinical decision making and care delivery (Koh et al., 2023). Design features in some AI applications such as robotics could impact patient dignity such as care robots infantilising older adults (Wong et al., 2024).
3.2 The accountability - authority - capability gap
Accountability for AI-mediated decisions may exceed nurses and midwives authority and capability especially if the professions lack education and training on AI, the autonomy to use it, or are excluded from AI development, procurement and implementation (Ronquillo et al., 2021). Overreliance on AI may also erode clinical judgement and make any capability deficits among nurses and midwives worse, creating a professional vulnerability. These issues could be compounded by legal ambiguity and poor governance frameworks (McDonald, 2024) as nurses and midwives ability to advocate for patients depends on their ability to interpret and document AI outputs and the authority to challenge them through formal governance processes (Penner et al., 2025).
3.3 Governance gaps and solutions
Governance structures may become fragmented with the introduction of AI, with uncertainty about who is responsible for AI-related decisions and care delivery (Alzani, 2023). This may leave frontline nursing and midwifery staff to bear any risks when AI is introduced despite their limited input into its development or implementation (O'Connor et al., 2023). Clear governance frameworks for AI in healthcare that define responsibility across clinical, administrative, and managerial roles with input multi-stakeholder input are needed (Almara, 2023; Luxton, 2014).
Theme 4: Beyond transparecy
Privacy, disclosure and earning trust
4.1 Privacy threats and relational harm
Privacy violations due to AI such as data breaches and surveillance may undermine trust and disrupt therapeutic relationships with patients and families (Adeyemo et al., 2025), and these risks may be amplified for nurses and midwives in resource-constrained settings (McDonald, 2024).
4.2 Building trust requires more than transparency
Placing trust in AI models and tools goes beyond transparency with how they work to include reliability, governance and professional validation (Dlugatch et al., 2024), with unverifiable sources likely to undermine nurses and midwives confidence in AI outputs (Chan et al., 2023). Therefore, professional involvement in AI systems development, implementation and ongoing evaluation could reinforce legitimacy and trust in AI tools (Issa et al., 2024).
4.3 Nurses and midwives' information governance roles
Nurses and midwives are central to translating AI outputs to patients and families including their limitations and risks to help maintain trust as well as integrating the outputs within care (Dreisbach et al., 2025). The professions may also hold stewardship roles, helping to verify the accuracy of AI model inputs and critically assessing AI outputs and biases (Penner et al., 2025).
Theme 5: Justice in development and deployment
From bias to equitable implementation
5.1 Algorithmic bias as structural inequality
Nurses and midwives worry about algorithmic bias in AI models and outputs especially where clinical documentation is concerned (Dell, 2025; Tang et al., 2024). AI bias could contribute to inequities in healthcare data, systems and decision making which could impact patient care and staff wellbeing (O'Connor & Booth, 2022)
5.2 Systematic exclusion of nurses and midwives
Nurses and midwives have limited input into AI design, development, procurement and governance although they may contribute to late stage validation and testing (Adeyemo et al., 2025; von Gerich et al., 2022). Marginalising their expertise could lead to incomplete or misaligned datasets, AI models and tools in healthcare.
5.3 Inequitable AI deployment concentrates risks
Deployment risks with AI include financial constraints which may limit the development, access and use of AI models and tools among nurses and midwives. This could exacerbate health inequalities by concentrating harms in vulnerable populations and low-resource settings (Koh et al., 2023) or vulnerable groups may face compounded risks if AI systems lack contextual validation or applied without adaptation to specific needs (Elgin & Elgin, 2024).
5.4 Equitable implementation
Equity throughout the AI lifecycle requires early consultation and active participation of all stakeholders including nurses and midwives for successful implementation to be achieved (De Raeve et al., 2021). Contextual adaptation of AI models and tools is essential to support care across culturally diverse populations and settings (Almagharbeh et al., 2025; Ibrahim et al., 2025).
Theme 6: Workforce preparation and deployment challenges
From education to practice integration
6.1 Capacity gaps and training requirements
AI literacy among the nursing and midwifery workforce is low with the professions having limited knowledge and skills. This is being exacerbated by the rapid pace of AI development moving from predictive, to generative, to agentic forms of AI within a few years which limits nurse educators abilities to upskill in AI and create contemporary curricula to educate students and practitioners about AI models and tools (O'Connor, 2022; Srinivasan et al., 2024).
6.2 Critical thinking and ethical awareness
Critical thinking and clinical judgement skills are needed by nurses and midwives when using AI tools to ensure their outputs are evaluated rather than just accepted (Hassan & El-Ashry, 2024). Ethical awareness and education amongst the professions is needed as they are responsible for contextualising AI outputs and safeguarding professional values and patient care (Migdadi et al., 2024; Woo et al., 2024).
6.3 Academic integrity and competency
Academic integrity is a concern in the era of AI as nursing and midwifery students have access to generative AI tools such as ChatGPT which may create challenges in learning and assessment if used inappropriately (Summers et al., 2024). The use of AI models and tools could also introduce fabricated and unverifiable outputs in research such as non-existent references and biased results (Woo et al., 2024). Robust governance frameworks are required in higher education and research to ensure responsible AI use by nurses and midwives (Tang et al., 2024).
6.4 Implementation barriers
Increased or disrupted workloads associated with AI system use could impact nurses and midwives who are also concerned about potential job displacement from AI tools (Hassan & El-Ashry, 2024). This may lead to resistance and anxiety related to AI adoption and use amongst the professions which may be compounded by inadequate organisational support e.g., insufficient technical infrastructure, unclear governance processes, lack of change management strategies (Ramadan et al., 2024). AI implementation needs to align with clinical, administrative and managerial workflows of nurses and midwives and needs to be facilitated by investment in IT systems and management support to engage the workforce and protect professional practice (von Gerich et al., 2022).
