<?xml version="1.0" encoding="UTF-8"?><?xml-stylesheet href="/in/rss-styles.xsl" type="text/xsl"?><rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hospital AI Lab</title><description>Columns on the practical side of medical AI in Ireland: HIQA’s national guidance, the EU AI Act, GDPR and patient data, device safety, and what hospital teams should ask vendors.</description><link>https://hospital-ai-lab.com/in/</link><language>en</language><atom:link href="https://hospital-ai-lab.com/in/rss.xml" rel="self" type="application/rss+xml"/><lastBuildDate>Mon, 16 Nov 2026 00:00:00 GMT</lastBuildDate><item><title>Hospital chatbots and AI-written text: what Article 50 asks</title><link>https://hospital-ai-lab.com/in/blog/2026-11-16-eu-ai-act-article-50-hospital-chatbot-ai-written-text/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-16-eu-ai-act-article-50-hospital-chatbot-ai-written-text/</guid><description>Does a patient know they are chatting with an AI? Article 50 of the EU AI Act sets disclosure rules for chatbots and AI-written content. What it says.</description><pubDate>Mon, 16 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;A hospital’s website now has a chat window that answers questions about appointments. A patient types, “Is this a real person?” The EU AI Act has a short article on exactly this: Article 50, “Transparency obligations for providers and deployers of certain AI systems”. This post reads the official English text of Regulation (EU) 2024/1689 as published in the Official Journal in July 2024. We have not read later EU amendments or any Irish implementing law, so check the current text before you rely on it. This is general information, not legal advice.&lt;/p&gt;
&lt;h2 id=&quot;the-chatbot-rule-and-who-it-is-aimed-at&quot;&gt;The chatbot rule, and who it is aimed at&lt;/h2&gt;
&lt;p&gt;Article 50(1) puts the duty on providers, the companies that build and place the system on the market. They must design AI systems intended to interact directly with people so that the people concerned are informed they are interacting with an AI system. The exception is where this is obvious to a reasonably well-informed, observant and circumspect person, given the circumstances and context of use. Article 50(5) adds that the information must be given in a clear and distinguishable way at the latest at the time of the first interaction, and must meet applicable accessibility requirements.&lt;/p&gt;
&lt;p&gt;A hospital that buys a chatbot is usually the deployer, not the provider. The article’s duty in this paragraph sits with the provider, so the practical step for the hospital is to ask the vendor how the product does this, and to check what a patient actually sees on screen.&lt;/p&gt;
&lt;h2 id=&quot;ai-generated-content&quot;&gt;AI-generated content&lt;/h2&gt;
&lt;p&gt;Article 50(2) asks providers of AI systems that generate synthetic audio, image, video or text to make sure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, as far as technically feasible. The duty does not apply to the extent the system performs an assistive function for standard editing or does not substantially alter the input data provided by the deployer or its meaning.&lt;/p&gt;
&lt;p&gt;Article 50(4) has a rule for deployers. Those using an AI system that generates or manipulates text published to inform the public on matters of public interest must disclose that the text was artificially generated or manipulated. This does not apply where the content has undergone human review or editorial control and a person holds editorial responsibility for publication. A hospital newsletter or public health notice written with AI could fall in this area, so it is worth deciding who holds editorial responsibility before you publish.&lt;/p&gt;
&lt;h2 id=&quot;what-the-article-does-not-settle&quot;&gt;What the article does not settle&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Article 50(6) says these paragraphs do not affect the requirements in Chapter III and are without prejudice to other transparency duties in Union or national law. Other rules, for example on personal data, may also apply to a chatbot that handles patient details.&lt;/li&gt;
&lt;li&gt;We have not read any guidance on how Article 50 applies in a hospital. The questions below are our suggestions, not requirements.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;questions-to-ask-before-the-chat-window-goes-live&quot;&gt;Questions to ask before the chat window goes live&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;Does the first message say, in plain words, that the patient is talking to an AI?&lt;/li&gt;
&lt;li&gt;Can the patient reach a person, and is that route easy to find?&lt;/li&gt;
&lt;li&gt;If staff use AI to draft public notices, who reviews them and who is the named editor?&lt;/li&gt;
&lt;li&gt;What does the vendor say about marking AI-generated output, and in what form?&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Article 50(1): providers must design chatbots so people are told they are interacting with an AI, unless it is obvious; the information comes at the latest at the first interaction (Article 50(5)).&lt;/li&gt;
&lt;li&gt;Article 50(2) and (4) cover marking AI-generated content and disclosing AI-written public-interest text, with an exception for human editorial control.&lt;/li&gt;
&lt;li&gt;A hospital that buys a tool should ask the vendor how it meets Article 50 and check what patients see. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Source: &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689 (Artificial Intelligence Act), Official Journal, 12 July 2024, Article 50&lt;/a&gt; (text read from the EU Publications Office). Read on 16 November 2026.&lt;/em&gt;&lt;/p&gt;
&lt;hr&gt;
&lt;p&gt;&lt;em&gt;This site’s content is for general information, research, and education. It does not replace medical diagnosis or advice for any specific patient, and it is not legal advice.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>Column</category></item><item><title>AI for Care 2026-2030: what hospital leaders should take from it</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-ai-for-care-strategy-six-principles-hospital-leaders/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-ai-for-care-strategy-six-principles-hospital-leaders/</guid><description>Ireland’s national AI strategy for healthcare sets four pillars and six principles. Here is what it means for the AI proposals that land on your desk.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;When a vendor says its product “fits the national strategy”, it helps to know what the strategy actually says. The Department of Health’s AI for Care strategy for 2026-2030 is the Irish document to read. It is a policy strategy, not legislation, and the copy we read carries no publication date. Its foreword is signed by the Minister for Health, Jennifer Carroll MacNeill.&lt;/p&gt;
&lt;h2 id=&quot;four-pillars&quot;&gt;Four pillars&lt;/h2&gt;
&lt;p&gt;The strategy sets four strategic pillars: AI for clinical care, AI for operations, AI for research and innovation, and AI for public health. For a hospital administrator the operations pillar is the one closest to daily work, but the same governance questions apply to all four.&lt;/p&gt;
&lt;h2 id=&quot;six-guiding-principles&quot;&gt;Six guiding principles&lt;/h2&gt;
&lt;p&gt;These are the strategy’s principles, which it says must be referenced throughout all AI efforts:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Person-centred:&lt;/strong&gt; health outcomes and the needs of the public are central to decisions about adopting AI.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Transparent and trustworthy:&lt;/strong&gt; people are told how, when and where AI is used to support healthcare decisions and what impact it might have.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Human in the loop:&lt;/strong&gt; AI supports clinicians, who stay at the centre of care. The strategy also commits to building AI literacy across the health service and among patients.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Lived experience:&lt;/strong&gt; a continuous learning approach that involves patients and clinicians.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Governance and safety:&lt;/strong&gt; AI operates within governance that aligns with health policy, regulation and ethics, including data protection requirements and AI-specific legislation.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Proven benefit:&lt;/strong&gt; a commitment to measurable value through monitoring and evaluation.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;how-it-connects-to-other-documents&quot;&gt;How it connects to other documents&lt;/h2&gt;
&lt;p&gt;The strategy says the HSE has developed a corresponding AI Implementation Framework, and that HIQA’s forthcoming national guidance will support compliance with the EU AI Act. HIQA has since published that guidance (29 July 2026). We have not read the HSE framework.&lt;/p&gt;
&lt;h2 id=&quot;using-it-in-a-vendor-meeting&quot;&gt;Using it in a vendor meeting&lt;/h2&gt;
&lt;p&gt;Turn the principles into questions:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Transparent:&lt;/strong&gt; how will patients be told, and who explains it?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Human in the loop:&lt;/strong&gt; who can override, and are they trained?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Governance and safety:&lt;/strong&gt; which rules apply, and where is the record?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Proven benefit:&lt;/strong&gt; what will we measure, and who looks at it?&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;A strategy is not a compliance certificate. Use it to frame a proposal, and use the binding documents (the AI Act, the GDPR, the Medical Device Regulation) to test it.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;AI for Care 2026-2030 sets four pillars and six guiding principles; it is a strategy, not a set of rules.&lt;/li&gt;
&lt;li&gt;It stresses keeping a human in the loop, transparency and AI literacy.&lt;/li&gt;
&lt;li&gt;Use it to frame vendor questions, not to prove compliance. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://www.drugsandalcohol.ie/46031/1/AI_for_Care_2026-2030.pdf&quot;&gt;Department of Health, AI for Care: The AI Strategy for Healthcare in Ireland 2026-2030 (PDF)&lt;/a&gt; (foreword and sections 4.2 and 5); &lt;a href=&quot;https://www.hiqa.ie/ga/node/74973&quot;&gt;HIQA press release (29 July 2026)&lt;/a&gt;. Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>Column</category></item><item><title>Before staff use an AI chatbot: the DPC’s questions for hospitals</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-dpc-ai-llm-guidance-questions-before-using-ai-product/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-dpc-ai-llm-guidance-questions-before-using-ai-product/</guid><description>Is a clinician pasting notes into a chatbot? The Data Protection Commission’s AI guidance lists what to understand first, and the risks it names, in plain language.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;Hospitals now face two kinds of AI use: the tools the organisation buys, and the chatbots staff open on their own. The Data Protection Commission (DPC) addressed both in “AI, Large Language Models and Data Protection”, published on 18 July 2024. It is not about hospitals specifically, but nearly every question in it applies to them.&lt;/p&gt;
&lt;h2 id=&quot;it-starts-with-a-warning&quot;&gt;It starts with a warning&lt;/h2&gt;
&lt;p&gt;The DPC says there may be a tendency to use large amounts of personal data in training, sometimes unnecessarily and without people’s knowledge. It also flags accuracy and retention problems, models shared with others for purposes people do not agree with, biased outputs, and fine-tuning that exposes people to risks where there were none before. Where personal data is involved, the GDPR applies to organisations using AI systems as well as to the providers.&lt;/p&gt;
&lt;h2 id=&quot;what-to-understand-before-using-an-ai-product&quot;&gt;What to understand before using an AI product&lt;/h2&gt;
&lt;p&gt;The guidance says an organisation using an AI product could be a data controller, and that it should consider a formal risk assessment. Before starting, it should understand:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;what personal data the system uses and how it uses it;&lt;/li&gt;
&lt;li&gt;where the data goes when a third party is involved;&lt;/li&gt;
&lt;li&gt;whether the provider retains it or re-uses it; and&lt;/li&gt;
&lt;li&gt;how the product lets the organisation meet its GDPR obligations. The documentation should tell you in an understandable form.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;risks-the-dpc-names&quot;&gt;Risks the DPC names&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Memorisation:&lt;/strong&gt; some models can regurgitate passages of training data unintentionally.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Filters can be circumvented,&lt;/strong&gt; so data you thought was blocked may appear.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Inaccurate or biased output&lt;/strong&gt; relied upon without critical human review creates automated decision-making risks.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Data subject rights:&lt;/strong&gt; if someone asks for access to or deletion of their data in the AI system, can you do it?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Storage limitation:&lt;/strong&gt; without a retention schedule, the organisation risks non-compliance.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Third-party products&lt;/strong&gt; can raise further security risks from what employees input.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;what-to-do-on-monday&quot;&gt;What to do on Monday&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;Say in writing which AI tools staff may use with patient information, and which they may not.&lt;/li&gt;
&lt;li&gt;For each approved tool, record the four things above.&lt;/li&gt;
&lt;li&gt;Make sure someone can handle access and deletion requests that touch the tool.&lt;/li&gt;
&lt;li&gt;Add a retention rule.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;If your hospital builds its own AI product using personal data, the guidance also asks you to consider whether there is a less risky non-AI route, to check that the purpose fits the legal basis you already have, and to think about an impact assessment.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;The DPC’s July 2024 guidance applies to organisations that use AI products and to those that build them.&lt;/li&gt;
&lt;li&gt;Understand what data the tool uses, where it goes, whether it is kept or re-used, and how you meet your GDPR duties.&lt;/li&gt;
&lt;li&gt;Watch for memorisation, circumvented filters, unreviewed outputs, rights requests and retention. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://www.dataprotection.ie/en/dpc-guidance/blogs/AI-LLMs-and-Data-Protection&quot;&gt;Data Protection Commission, AI, Large Language Models and Data Protection (18 July 2024)&lt;/a&gt;. Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>News</category></item><item><title>EU AI Act: are we a deployer, and what must the hospital do?</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-eu-ai-act-hospital-deployer-duties-article-26/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-eu-ai-act-hospital-deployer-duties-article-26/</guid><description>If your hospital uses a vendor’s AI system, the EU AI Act gives you duties of your own. Article 26 in plain language, with the dates as published.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;Many hospitals assume the AI Act is the vendor’s problem. It is not entirely. The Regulation (EU) 2024/1689 applies directly in Ireland, and it gives separate duties to “deployers”: anyone using an AI system under their authority, except in a personal non-professional activity. A hospital using a vendor’s AI product is a deployer.&lt;/p&gt;
&lt;h2 id=&quot;when-does-it-apply&quot;&gt;When does it apply?&lt;/h2&gt;
&lt;p&gt;As published in the Official Journal on 12 July 2024, Article 113 says the Regulation applies from 2 August 2026, with exceptions: Chapters I and II (general provisions and prohibited practices) from 2 February 2025, and Article 6(1) with its corresponding obligations from 2 August 2027. We read the text as published; EU amendments after that were not read, so check the current version.&lt;/p&gt;
&lt;h2 id=&quot;which-systems-are-high-risk&quot;&gt;Which systems are “high-risk”?&lt;/h2&gt;
&lt;p&gt;Article 26 applies to high-risk systems. Two routes matter to hospitals. Under Article 6(1), an AI system that is a product, or a safety component of a product, covered by the EU product laws in Annex I (which include the Medical Device Regulation and the In Vitro Diagnostic Regulation), and that needs third-party conformity assessment, is high-risk. Under Annex III, point 5, uses include evaluating eligibility for public healthcare services and emergency healthcare patient triage systems.&lt;/p&gt;
&lt;h2 id=&quot;what-a-deployer-must-do-article-26&quot;&gt;What a deployer must do (Article 26)&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Use the system according to its &lt;strong&gt;instructions for use&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;Assign &lt;strong&gt;human oversight&lt;/strong&gt; to people with the necessary competence, training, authority and support.&lt;/li&gt;
&lt;li&gt;Where you control the input data, make sure it is &lt;strong&gt;relevant and sufficiently representative&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Monitor&lt;/strong&gt; the system. If you think it may present a risk, inform the provider or distributor and the market surveillance authority, and suspend use.&lt;/li&gt;
&lt;li&gt;For a &lt;strong&gt;serious incident&lt;/strong&gt;, inform the provider first, then the importer or distributor and the authorities.&lt;/li&gt;
&lt;li&gt;Keep the &lt;strong&gt;logs&lt;/strong&gt; you control for at least six months, unless other law says otherwise.&lt;/li&gt;
&lt;li&gt;Before using a high-risk system at work, &lt;strong&gt;tell workers&lt;/strong&gt; and their representatives.&lt;/li&gt;
&lt;li&gt;Use the provider’s information to help with a &lt;strong&gt;data protection impact assessment&lt;/strong&gt; where applicable.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;what-to-do-now&quot;&gt;What to do now&lt;/h2&gt;
&lt;p&gt;Ask each vendor three things in writing: do you consider this product high-risk, and under which provision; what do your instructions for use say about oversight; and what logs does the system keep and who controls them. Then name an overseer and put the incident route on a card.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;A hospital that uses an AI system under its authority is a deployer with its own duties.&lt;/li&gt;
&lt;li&gt;Article 26 covers instructions, oversight, input data, monitoring, incidents, logs, informing workers and impact assessments.&lt;/li&gt;
&lt;li&gt;Dates are as published: general application 2 August 2026, Article 6(1) from 2 August 2027. Check for amendments. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689 (AI Act), Official Journal&lt;/a&gt; (Articles 6, 26 and 113, Annex I Section A, Annex III point 5). Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item><item><title>AI literacy: does the EU AI Act require us to train hospital staff?</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-eu-ai-act-ai-literacy-article-4-hospital-staff-training/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-eu-ai-act-ai-literacy-article-4-hospital-staff-training/</guid><description>Article 4 of the AI Act has applied since February 2025. What it asks of a hospital that uses AI, and a simple way to start a training plan.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;A nurse manager asks the question after a vendor demo: “Are we supposed to train everyone on this?” The EU AI Act has a short answer in Article 4, and it has been in force for longer than most of the Act.&lt;/p&gt;
&lt;h2 id=&quot;what-article-4-says&quot;&gt;What Article 4 says&lt;/h2&gt;
&lt;p&gt;Providers and deployers of AI systems shall take measures to ensure, to their best extent, a sufficient level of AI literacy of their staff and other persons dealing with the operation and use of AI systems on their behalf. The measures should take into account their technical knowledge, experience, education and training, the context in which the systems are used, and the people on whom the systems are used.&lt;/p&gt;
&lt;p&gt;Three points stand out for a hospital:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;It applies to &lt;strong&gt;deployers&lt;/strong&gt;, which includes a hospital using a vendor’s system under its authority.&lt;/li&gt;
&lt;li&gt;It reaches &lt;strong&gt;other persons acting on your behalf&lt;/strong&gt;, not only employees. Think agency staff or contractors.&lt;/li&gt;
&lt;li&gt;It is scaled to &lt;strong&gt;context and role&lt;/strong&gt;: “to their best extent” and “taking into account” are the Act’s own words. It does not set a syllabus or an exam.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id=&quot;since-when&quot;&gt;Since when?&lt;/h2&gt;
&lt;p&gt;Article 113(a) says Chapters I and II of the Regulation apply from 2 February 2025, and Article 4 is in Chapter I. That is earlier than the general application date of 2 August 2026, as published. We did not read later EU amendments, so check the current text.&lt;/p&gt;
&lt;h2 id=&quot;the-irish-policy-angle&quot;&gt;The Irish policy angle&lt;/h2&gt;
&lt;p&gt;The Department of Health’s AI for Care strategy also commits to building AI literacy across the health service and among patients and service users, under its “human in the loop” principle. HIQA’s draft guidance names training under its responsiveness principle. Neither document sets a course; both treat literacy as part of using AI safely.&lt;/p&gt;
&lt;h2 id=&quot;a-simple-starting-plan&quot;&gt;A simple starting plan&lt;/h2&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;List who touches each AI tool:&lt;/strong&gt; clinicians, nurses, administrators, IT, front-desk staff and agency workers.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Write a one-page brief per tool:&lt;/strong&gt; what it is for, what it must not be used for, how to override it, and how to report a problem.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Match the depth to the role:&lt;/strong&gt; a nurse who reads its alerts needs different material from the person who configures it.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Record it:&lt;/strong&gt; who was briefed and when, and re-brief after significant updates.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;This plan is the author’s suggestion, not a requirement of Article 4.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Article 4 asks deployers to ensure a sufficient level of AI literacy among staff and others who operate or use AI systems for them.&lt;/li&gt;
&lt;li&gt;It applies from 2 February 2025 as published, and it is scaled to knowledge, role and context.&lt;/li&gt;
&lt;li&gt;Ireland’s AI for Care strategy and HIQA’s draft guidance point the same way. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689 (AI Act), Official Journal&lt;/a&gt; (Articles 4 and 113); &lt;a href=&quot;https://www.drugsandalcohol.ie/46031/1/AI_for_Care_2026-2030.pdf&quot;&gt;Department of Health, AI for Care 2026-2030 (PDF)&lt;/a&gt; (section 4.2). Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item><item><title>New AI tool, patient data: when do we need an impact assessment?</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-gdpr-impact-assessment-ai-tool-patient-data/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-gdpr-impact-assessment-ai-tool-patient-data/</guid><description>Before an AI tool goes live on patient data, GDPR Article 35 may require a data protection impact assessment. When, who is involved, and the 72-hour breach clock.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;“Do we need a DPIA for this?” is one of the first questions a project team asks the data protection officer. The honest answer sits in two articles of the GDPR, and in a short note from Ireland’s Data Protection Commission.&lt;/p&gt;
&lt;h2 id=&quot;why-patient-data-is-special&quot;&gt;Why patient data is special&lt;/h2&gt;
&lt;p&gt;Article 9(1) places data concerning health among the special categories of personal data whose processing is prohibited unless a condition in Article 9(2) applies. Condition (h) covers processing necessary for medical diagnosis, the provision of health or social care or treatment, or the management of health or social care systems and services, on the basis of Union or Member State law or under a contract with a health professional. Article 9(3) adds that such data should be processed by or under the responsibility of a professional subject to professional secrecy.&lt;/p&gt;
&lt;h2 id=&quot;when-an-impact-assessment-is-required&quot;&gt;When an impact assessment is required&lt;/h2&gt;
&lt;p&gt;Article 35(1) says that where processing, in particular using new technologies, is likely to result in a high risk to people’s rights and freedoms, the controller must carry out an assessment before the processing. Article 35(3)(b) names processing on a large scale of special categories of data. A hospital rolling out an AI tool across many patients’ records is close to that description, though whether a given project meets it is a judgement for your data protection officer. Article 35(2) asks the controller to seek the officer’s advice.&lt;/p&gt;
&lt;p&gt;The Data Protection Commission’s guidance on AI says an assessment may be needed, especially where the technology or processing is new to the organisation, where data sets are combined or where data about vulnerable people is used. It adds that doing one can be good practice anyway.&lt;/p&gt;
&lt;h2 id=&quot;a-link-to-the-ai-act&quot;&gt;A link to the AI Act&lt;/h2&gt;
&lt;p&gt;If the tool is a high-risk AI system under the AI Act, Article 26(9) says deployers use the information the provider gives under Article 13 to help carry out their GDPR impact assessment where applicable. Ask your vendor for that information early.&lt;/p&gt;
&lt;h2 id=&quot;the-breach-clock&quot;&gt;The breach clock&lt;/h2&gt;
&lt;p&gt;If something goes wrong, Article 33(1) asks the controller to notify the supervisory authority without undue delay and, where feasible, within 72 hours of becoming aware, unless the breach is unlikely to result in a risk to people’s rights and freedoms. The processor must tell the controller without undue delay (Article 33(2)). Write the vendor’s alert time, in hours, into the contract.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Health data is special category data; Article 35 requires an impact assessment before high-risk processing, naming large-scale use of special category data.&lt;/li&gt;
&lt;li&gt;Involve your data protection officer, and use the provider’s information for the assessment.&lt;/li&gt;
&lt;li&gt;A personal data breach has a 72-hour clock for the controller. This is general information, not legal advice; Irish statutes were not read for this post.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2016/679/oj&quot;&gt;Regulation (EU) 2016/679 (GDPR)&lt;/a&gt; (Articles 9, 33 and 35); &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689, Article 26(9)&lt;/a&gt;; &lt;a href=&quot;https://www.dataprotection.ie/en/dpc-guidance/blogs/AI-LLMs-and-Data-Protection&quot;&gt;Data Protection Commission, AI, Large Language Models and Data Protection (18 July 2024)&lt;/a&gt;. Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item><item><title>HIQA’s national AI guidance: what should your hospital set up?</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-hiqa-national-ai-guidance-hospital-governance-checklist/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-hiqa-national-ai-guidance-hospital-governance-checklist/</guid><description>Is your hospital using AI without a list of tools or a named owner? HIQA’s guidance and its January draft point to five things to put in place, in plain language.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;On 29 July 2026 HIQA published Ireland’s first National Guidance for the Responsible and Safe Use of Artificial Intelligence in Health and Social Care Services. If you run or advise a hospital, it is the most practical Irish document on the subject so far. One caveat up front: we could not open the final PDF, so this post relies on HIQA’s press release and on the consultation draft from January 2026. Where a detail comes from the draft, we say so, because the final text may differ.&lt;/p&gt;
&lt;h2 id=&quot;what-hiqa-says-it-is&quot;&gt;What HIQA says it is&lt;/h2&gt;
&lt;p&gt;HIQA’s press release says the guidance was commissioned by the Department of Health and developed with the Department and the HSE. It is intended for all services providing health and social care in Ireland, and it is also designed to educate and empower people using services about what to expect when AI is used in their care. It is organised around HIQA’s four principles for person-centred care and support: accountability, a human rights-based approach, safety and wellbeing, and responsiveness. It builds on HIQA’s existing national standards. It is guidance, not a statute.&lt;/p&gt;
&lt;h2 id=&quot;five-things-the-draft-suggests-a-service-puts-in-place&quot;&gt;Five things the draft suggests a service puts in place&lt;/h2&gt;
&lt;p&gt;The January 2026 draft gives examples under its accountability principle. They read like a project checklist:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;A list of authorised AI tools,&lt;/strong&gt; kept by the service and shared with staff.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;A named person&lt;/strong&gt; who compiles and maintains an inventory of AI tools in use.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Expert input before first use:&lt;/strong&gt; legal and regulatory advice, including on ethical and cyber security impacts and EU AI Act requirements.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Quality control and risk assessment:&lt;/strong&gt; assurance practices for AI tools, with risk assessments inside existing risk management procedures.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Fundamental rights impact assessments&lt;/strong&gt; for high-risk tools, as the EU AI Act requires.&lt;/li&gt;
&lt;/ol&gt;
&lt;p&gt;The draft adds that structures should be proportionate to the size and complexity of the service. A small clinic does not need a hospital’s committee.&lt;/p&gt;
&lt;h2 id=&quot;people-stay-in-charge&quot;&gt;People stay in charge&lt;/h2&gt;
&lt;p&gt;According to the draft, AI tools should augment, not replace, human judgement and must sit inside a process with appropriate human oversight. It describes a human who oversees decisions in real time (“human in the loop”) or a system that can be overridden (“human on the loop”), and says staff should be able to override an output. Management is responsible for decisions on using and monitoring AI tools, staff for interpreting outputs, and clinical staff remain responsible for their clinical decisions.&lt;/p&gt;
&lt;h2 id=&quot;if-you-build-it-yourself&quot;&gt;If you build it yourself&lt;/h2&gt;
&lt;p&gt;The draft also warns that medical device software, including AI-based software, may be developed by hospitals themselves. In that case it says the hospital would need to meet provider obligations under the EU AI Act and manufacturer obligations under the Medical Device Regulation. Ask your IT and clinical engineering teams whether anything in-house could fall under that.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;HIQA’s national AI guidance (29 July 2026) is for all Irish health and social care services and is built on four principles.&lt;/li&gt;
&lt;li&gt;The January 2026 draft suggests a list of AI tools, a named inventory owner, expert input, risk assessments and fundamental rights assessments for high-risk tools.&lt;/li&gt;
&lt;li&gt;People stay responsible and able to override. In-house software may make you a provider or manufacturer.&lt;/li&gt;
&lt;li&gt;This is general information, not legal advice. Read the final guidance and ask your compliance officer.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://www.hiqa.ie/ga/node/74973&quot;&gt;HIQA, press release on the National Guidance (29 July 2026)&lt;/a&gt;; &lt;a href=&quot;https://www.hiqa.ie/sites/default/files/2026-01/Draft-National-Guidance-for-the-Responsible-and-Safe-Use-of-Artificial-Intelligence-in-Health-and-Social-Care-Services-for-public-consultation.pdf&quot;&gt;HIQA, Draft National Guidance for the Responsible and Safe use of AI in Health and Social Care Services (January 2026, PDF)&lt;/a&gt; (Principle 1: Accountability; sections 2.2 and 2.4). Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item><item><title>AI device fault or software update: how to report to HPRA</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-hpra-device-incident-reporting-software-update-field-safety-notice/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-hpra-device-incident-reporting-software-update-field-safety-notice/</guid><description>What counts as a serious incident, how a hospital reports it to HPRA, and a six-step routine for field safety notices that arrive as software updates.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;At 2am on a ward, an infusion pump or a monitoring app misbehaves after an update. Who does the nurse in charge tell, and what happens next? HPRA’s pages for hospitals answer both questions, and the Medical Device Regulation explains the manufacturer’s side.&lt;/p&gt;
&lt;h2 id=&quot;what-hpra-asks-hospitals-to-report&quot;&gt;What HPRA asks hospitals to report&lt;/h2&gt;
&lt;p&gt;HPRA says hospitals and healthcare professionals should report serious incidents. A serious incident is any incident that directly or indirectly led, might have led or might lead to the death of a patient, user or other person, the temporary or permanent serious deterioration of someone’s state of health, or a serious public health threat. If you are in doubt, HPRA encourages you to report anyway.&lt;/p&gt;
&lt;p&gt;You can report through HPRA’s online report form, by emailing a completed incident report form to &lt;a href=&quot;mailto:devicesafety@hpra.ie&quot;&gt;devicesafety@hpra.ie&lt;/a&gt;, by post to the Medical Devices Department, or by phone on +353 (0)1 676 4971. HPRA’s page sets no deadline for hospitals, so decide your own internal target and write it on the incident card.&lt;/p&gt;
&lt;h2 id=&quot;the-manufacturers-clocks&quot;&gt;The manufacturer’s clocks&lt;/h2&gt;
&lt;p&gt;Under Article 87 of the MDR, manufacturers report a serious incident immediately after establishing a causal relationship, and not later than 15 days after becoming aware. For a serious public health threat the limit is 2 days, and for death or an unanticipated serious deterioration in health it is 10 days. Article 87(10) asks Member States to encourage and enable healthcare professionals, users and patients to report suspected serious incidents. These are the manufacturer’s deadlines, not the hospital’s.&lt;/p&gt;
&lt;h2 id=&quot;when-the-notice-comes-from-the-manufacturer&quot;&gt;When the notice comes from the manufacturer&lt;/h2&gt;
&lt;p&gt;A manufacturer may send a Field Safety Notice (FSN) describing a Field Safety Corrective Action (FSCA). HPRA lists software updates among examples, alongside changes to instructions for use, repair, recall or replacement. HPRA says each hospital manages these under its own quality or risk management protocols, and recommends six steps:&lt;/p&gt;
&lt;ol&gt;
&lt;li&gt;&lt;strong&gt;Receive and notify:&lt;/strong&gt; a documented process to receive notices and pass them to clinical, technical and procurement staff.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Assess applicability:&lt;/strong&gt; do we hold affected devices? Traceability is essential.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Assess risk:&lt;/strong&gt; if devices have been used, are patients at risk or in need of follow-up?&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Implement&lt;/strong&gt; the manufacturer’s instructions.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Document&lt;/strong&gt; receipt, devices identified, actions, dates and people, and return acknowledgements.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Report&lt;/strong&gt; incidents to HPRA where required.&lt;/li&gt;
&lt;/ol&gt;
&lt;h2 id=&quot;where-ai-fits&quot;&gt;Where AI fits&lt;/h2&gt;
&lt;p&gt;If the AI-enabled product is a device, these routes apply to its software updates too. If it is a high-risk AI system under the AI Act, a deployer also informs the provider first and then the authorities of a serious incident (Article 26(5)).&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;HPRA asks hospitals to report serious incidents (death, serious deterioration in health, serious public health threat), and encourages reporting if in doubt.&lt;/li&gt;
&lt;li&gt;Manufacturers have 15, 10 or 2-day clocks under the MDR; HPRA sets none for hospitals, so set your own.&lt;/li&gt;
&lt;li&gt;Treat software updates as possible field safety actions, with a six-step routine. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://www.hpra.ie/regulation/medical-devices/health-institutions-and-healthcare-professionals/reporting-medical-device---ivd-safety-issues&quot;&gt;HPRA, Reporting medical device &amp;#x26; IVD safety issues&lt;/a&gt;; &lt;a href=&quot;https://www.hpra.ie/regulation/medical-devices/health-institutions-and-healthcare-professionals/managing-field-safety-corrective-actions&quot;&gt;HPRA, Managing field safety corrective actions&lt;/a&gt;; &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2017/745/oj&quot;&gt;Regulation (EU) 2017/745, Article 87&lt;/a&gt;; &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689, Article 26(5)&lt;/a&gt;. Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item><item><title>Is our hospital AI software a medical device? MDR rule 11 explained</title><link>https://hospital-ai-lab.com/in/blog/2026-11-15-is-hospital-ai-software-a-medical-device-mdr-rule-11-hpra/</link><guid isPermaLink="true">https://hospital-ai-lab.com/in/blog/2026-11-15-is-hospital-ai-software-a-medical-device-mdr-rule-11-hpra/</guid><description>A vendor says its AI is “not a medical device”. How the MDR defines software, how rule 11 sets the class, and what HPRA does and does not do in Ireland.</description><pubDate>Sun, 15 Nov 2026 00:00:00 GMT</pubDate><content:encoded>&lt;p&gt;“It’s just software, not a medical device.” Hospitals hear this often. Whether it is true depends on what the manufacturer intends the software to do, not on whether it sits in a box. The Medical Device Regulation (MDR, Regulation (EU) 2017/745) sets the test. We read the text as published in the Official Journal in 2017; consolidated versions may differ.&lt;/p&gt;
&lt;h2 id=&quot;the-definition-names-software&quot;&gt;The definition names software&lt;/h2&gt;
&lt;p&gt;Article 2(1) defines a medical device as any instrument, apparatus, appliance, &lt;strong&gt;software&lt;/strong&gt;, implant, reagent, material or other article intended by the manufacturer to be used for human beings for specific medical purposes. The list of purposes includes diagnosis, prevention, monitoring, prediction, prognosis, treatment or alleviation of disease. So the question for your vendor is: what is the intended purpose, in writing, and does our planned use match it?&lt;/p&gt;
&lt;h2 id=&quot;rule-11-sets-the-class-of-software&quot;&gt;Rule 11 sets the class of software&lt;/h2&gt;
&lt;p&gt;Annex VIII, rule 11, says:&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;Software that provides information used to take decisions with diagnosis or therapeutic purposes is &lt;strong&gt;class IIa&lt;/strong&gt;, except where the decisions could cause death or irreversible deterioration (&lt;strong&gt;class III&lt;/strong&gt;) or serious deterioration of health or a surgical intervention (&lt;strong&gt;class IIb&lt;/strong&gt;).&lt;/li&gt;
&lt;li&gt;Software that monitors physiological processes is class IIa, or IIb where it monitors vital parameters whose variations could result in immediate danger.&lt;/li&gt;
&lt;li&gt;All other software is class I.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;The class also matters for the AI Act: under Article 6(1), an AI system in a product covered by the MDR that must undergo third-party conformity assessment is high-risk, from 2 August 2027 as published.&lt;/p&gt;
&lt;h2 id=&quot;what-hpra-does&quot;&gt;What HPRA does&lt;/h2&gt;
&lt;p&gt;HPRA says it is Ireland’s National Competent Authority for the MDR and the IVDR. It monitors devices and IVDs on the Irish market and says it does not play a role in the authorisation or approval of devices. For a hospital that means the evidence of conformity comes from the manufacturer; ask for the CE-marking details.&lt;/p&gt;
&lt;h2 id=&quot;if-you-build-it-in-house&quot;&gt;If you build it in-house&lt;/h2&gt;
&lt;p&gt;Article 5(5) lets devices made and used only within health institutions escape most MDR requirements if conditions are met, including a justification that the target patient group’s needs cannot be met by an equivalent device on the market, a public declaration, documentation for the competent authority, and review of clinical experience. HPRA lists in-house manufacturing among the routes hospitals should understand. HIQA’s draft guidance also warns that a hospital building AI-based device software takes on provider and manufacturer obligations.&lt;/p&gt;
&lt;h2 id=&quot;key-takeaways&quot;&gt;Key takeaways&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;The MDR names software in its definition of a medical device; intended purpose decides.&lt;/li&gt;
&lt;li&gt;Rule 11 puts decision-support software in class IIa or higher; a vendor’s class should be explained in writing.&lt;/li&gt;
&lt;li&gt;HPRA monitors but does not approve; ask for conformity evidence. In-house devices have their own conditions. This is general information, not legal advice.&lt;/li&gt;
&lt;/ul&gt;
&lt;p&gt;&lt;em&gt;Sources: &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2017/745/oj&quot;&gt;Regulation (EU) 2017/745 (MDR), Official Journal&lt;/a&gt; (Articles 2(1) and 5(5), Annex VIII rule 11); &lt;a href=&quot;https://www.hpra.ie/regulation/medical-devices/health-institutions-and-healthcare-professionals/overview-of-medical-device---ivd-regulation-for-hospitals&quot;&gt;HPRA, Overview of medical device &amp;#x26; IVD regulation for hospitals&lt;/a&gt;; &lt;a href=&quot;https://eur-lex.europa.eu/eli/reg/2024/1689/oj&quot;&gt;Regulation (EU) 2024/1689, Article 6(1)&lt;/a&gt;. Read on 15 November 2026.&lt;/em&gt;&lt;/p&gt;</content:encoded><category>AI Tools</category></item></channel></rss>