HIQA Publishes First National Guidance For Safe AI Use In Health And Social Care

Ireland’s new national guidance says artificial intelligence should support not replace professional judgement, with services expected to establish clear governance, human oversight and safeguards for safety, privacy and equality.

The Health Information and Quality Authority (HIQA) has published Ireland’s first national guidance on the responsible and safe use of artificial intelligence in health and social care services.

Published on 29 July 2026, the guidance applies to organisations and staff across Irish health and social care. It is also intended to help patients, residents and service users understand what they should expect when an artificial intelligence tool is used as part of their care.

For nurses and midwives, the guidance has implications for clinical decision support, documentation, risk management, patient information, data protection and professional accountability.

It makes a central point clear: artificial intelligence can support professional practice, but it does not replace human judgement or transfer responsibility away from the healthcare professional making a clinical decision.

What the HIQA AI guidance covers

The guidance was commissioned by the Department of Health and developed by HIQA with the Department, the Health Service Executive and other health and social care stakeholders.

HIQA describes it as a good-practice guidance document rather than a replacement for legislation, professional regulation or local clinical-governance requirements.

It is organised around four principles:

1. Accountability
2. A human rights-based approach
3. Safety and wellbeing
4. Responsiveness

Under each principle, HIQA outlines practical considerations for organisations and staff during the planning, procurement, deployment, monitoring and eventual withdrawal of AI tools.

HIQA Director of Health Information and Standards Rachel Flynn said the guidance was intended to keep “patient safety, transparency and accountability” at the forefront as AI becomes more widely used.

Accountability remains with services and professionals

HIQA describes accountability as the foundation of responsible and safe healthcare AI.

Services should have formal governance structures, approved policies, clear lines of responsibility and oversight of every authorised AI tool. They should be able to explain who approved a tool, its permitted purpose, how its performance is monitored and what happens if it produces unreliable results.

Management is responsible for matters including procurement, governance, auditing and monitoring. Staff remain responsible for interpreting outputs and using tools in accordance with their professional obligations.

The guidance states that clinical staff remain responsible for their clinical decisions.

Healthcare professionals should understand, question and, where necessary, override an AI-generated recommendation when it conflicts with their professional judgement or the needs of the person receiving care.

How AI may enter nursing practice

HIQA identifies several current and potential applications of artificial intelligence in health and social care.

These include:

• Supporting the interpretation of radiology scans
• Helping to forecast admissions and bed occupancy
• Producing clinical decision-support recommendations
• Supporting continuous glucose monitoring
• Tailoring rehabilitation programmes
• Assisting with population-health screening
• Scheduling operating theatres and other resources
• Supporting clinical note-taking through ambient AI systems
• Providing information through virtual assistants or chatbots

These applications carry different levels of risk. An administrative scheduling tool does not necessarily require the same controls as a system that influences diagnosis, triage or treatment.

Services should therefore assess each tool according to its intended purpose, the population affected and the potential consequences of inaccurate or biased outputs.

Patient information and human rights

HIQA says services should explain when, where and why an AI tool is being used in a person’s care.

Information should be understandable and proportionate to the tool’s level of risk. People using services should be able to ask questions, seek clarification and participate in decisions affecting their care.

Where a patient or service user interacts directly with an AI chatbot or virtual assistant, the guidance says there should be a clearly signposted option to return to communication with a person.

The guidance uses the FREDA human-rights framework: fairness, respect, equality, dignity and autonomy.

Services should assess whether an AI tool could disadvantage particular groups because its training or validation data are not representative of the population in which it will be used.

Outputs should be monitored for bias, discrimination and unequal outcomes. Accessible and inclusive design should also be considered when tools are procured or developed.

Confidentiality and public AI tools

Personal health information entered into an AI system remains subject to data-protection, confidentiality and information-governance requirements.

HIQA says services must maintain strong data governance and comply with the General Data Protection Regulation and other relevant privacy and security legislation.

Its implementation checklist specifically calls for controls preventing identifiable information from being entered into publicly available large language models.

For frontline staff, this means that publicly accessible generative-AI services should not be treated as approved clinical systems. Patient names, clinical histories, photographs, recordings or other identifiable information must not be entered into an open AI service unless the organisation has expressly approved the system and put the necessary legal, security and governance arrangements in place.

Even apparently anonymous material may carry a risk of identification when combined with dates, locations, rare diagnoses or other contextual details.

Monitoring safety after an AI tool is introduced

Approval of a tool is not the end of the safety process.

HIQA says services should continuously monitor AI tools for reliability, accuracy, data quality, bias and changes in performance. This includes watching for “automation bias”, where staff accept an automated recommendation without sufficiently questioning it.

Clinical and corporate risk-management processes should be updated to cover AI. Tools should be entered on relevant risk registers, and services should be able to suspend their use if safety concerns arise.

Incident-management and open-disclosure processes should also cover harm or potential harm associated with AI.

Business-continuity plans are required so that care can continue safely if a system fails, becomes unavailable, is withdrawn or is affected by a cybersecurity incident.

Training and competence

The guidance says organisations should assess their workforce’s training needs before introducing an AI tool.

Training should be appropriate to each person’s role and should cover:

• The tool’s intended purpose
• Its evidence base and known limitations
• How to interpret and critically evaluate its outputs
• How to recognise bias or inaccurate information
• Where and how the tool stores information
• Relevant reporting and escalation arrangements
• The limits of the staff member’s authority and responsibility

Training should continue after implementation, rather than being confined to a single introductory session.

HIQA also warns against deskilling. Staff should retain the professional knowledge and practical ability needed to continue care safely if an AI system fails or is withdrawn.

Professional regulation still applies

HIQA notes that Ireland’s professional regulators, including the Nursing and Midwifery Board of Ireland, regulate the competence and conduct of individual practitioners.

Existing requirements concerning consent, confidentiality, competence, professional judgement, documentation and patient safety continue to apply when AI is used.

Professional regulators may issue more specific requirements concerning AI literacy, competence, supervision and scope of practice. Until then, nurses and midwives should treat AI as another clinical or information tool that must be used within their competence, organisational policy and professional code.

The legal position

The guidance sits alongside the European Union Artificial Intelligence Act, medical-device legislation, the General Data Protection Regulation and other Irish and European legal requirements.

The EU AI Act takes a risk-based approach. Certain medical-device software, emergency healthcare triage systems and other tools capable of significantly affecting health, safety or fundamental rights may be categorised as high risk.

HIQA notes that obligations under the EU AI Act are being introduced in stages. It also says the Regulation of Artificial Intelligence Bill 2026 is intended to establish Ireland’s national enforcement and regulatory structures.

The HIQA guidance should therefore be read as a practical governance resource, not as a complete statement of every legal obligation that may apply to a particular AI system.

HIQA says the role of digital technologies will be incorporated further as it updates and develops national standards.

The Department of Health, HSE and HIQA are also progressing related work, including the national AI for Care strategy and the HSE’s AI Implementation Framework.