Naoise Ó Cearúil (Kildare North, Fianna Fail)
Link to this: Individually | In context | Oireachtas source

283. To ask the Minister for Public Expenditure and Reform to outline the governance structures, risk assessment processes and ethical safeguards being considered as part of the Government's assessment of artificial intelligence deployment across the public sector; and if he will make a statement on the matter. [39799/26]

Jack Chambers (Dublin West, Fianna Fail)
Link to this: Individually | In context | Oireachtas source

My Department has taken a proactive and coordinated approach to supporting the responsible and effective adoption of Artificial Intelligence (AI) across the Public Service. A key component of this was the publication of the ‘Guidelines for the Responsible use of Artificial Intelligence in the Public Service’ in May 2025, which provide a practical framework to assist Public Service Bodies in designing, developing and deploying AI in a manner that is ethical, transparent and aligned with public service values.

The Guidelines are structured around seven core principles. These are (1) human agency and oversight, (2) technical robustness and safety, (3) privacy and data governance, (4) transparency, (5) diversity, non-discrimination and fairness, (6) societal and environmental well-being, and (7) accountability. These core principles are aligned with the European Commission’s High Level Expert Group’s principles for Trustworthy AI, and maintain a focus on ensuring use of AI is undertaken in an ethical manner. These are supported by a decision framework, an AI lifecycle model, and a practical Responsible AI Canvas tool. Together, these elements help Public Service Bodies to assess the appropriateness of AI use cases, manage risks, and embed good governance throughout the AI project lifecycle.

The second principle, technical robustness and safety, highlights that ensuring a full risk assessment being conducted is an essential part of using AI in a responsible and ethical manner. The Decision Framework can be used as a guide for public service workers when considering using AI to solve a problem or improve a service. This framework will help evaluate if AI is the most suitable solution. This framework is supported by the Responsible AI Canvas Tool which features a section for risk assessment and mitigation, ensuring these aspects are fully considered during the initial project stages. The AI Guidelines have received a high level of engagement, having had over 22,000 views and 14,797 downloads on the gov.ie AI resources webpage.

To build capability and ensure consistent application of these standards, my Department has worked with the Institute of Public Administration (IPA) to develop a suite of training courses, including Implementing the AI Guidelines, AI Essentials, Unlocking the Power of Microsoft Copilot for the Public Service, and an AI Masterclass for Senior Public Service Leaders. These courses are designed to meet the needs of both technical and non-technical staff and further details can be found on the IPA website www.ipa.ie. Since they were launched, very high uptake has been observed and over 5,500 people have completed the course on Navigating the Guidelines for the Responsible Use of Artificial Intelligence in the Public Service, and over 10,000 have completed the bespoke course on AI in the Public Service.

The Guidelines were also developed in line with the EU AI Act’s risk-based approach. The EU AI Act provides the overarching binding legal framework for the development and deployment of AI systems in the EU. The Act prohibits certain unacceptable-risk AI practices, and imposes obligations in respect of high-risk AI systems, transparency, governance and AI literacy.

The prohibitions on certain AI practices and AI literacy obligations have applied since 2 February 2025; and governance provisions and obligations for general-purpose AI models have applied since 2 August 2025. Article 27 of the EU AI Act introduces the fundamental rights impact assessment (FRIA) — a mandatory obligation for certain deployers of high-risk AI systems. The FRIA must cover the process in which the AI system will be used, duration & frequency of use, categories of persons or groups affected, specific risks of harm, human-oversight measures, as well as mitigation, governance and complaints arrangements. The obligation applies on first use, and the assessment must be updated if the relevant circumstances change. While this obligation was due to enter into force on 2 August 2026, the European Commission’s Digital Omnibus on AI package includes a proposal to extend the application timeline for the high-risk AI provisions of the AI Act to 2 December 2027 in respect of stand-alone high-risk AI systems falling under Article 6(2) and Annex III, and to 2 August 2028 in respect of high-risk AI systems embedded in products under Article 6(1) and Annex I.

The proposal is currently under negotiations at EU level, with provisional agreement reached on 7 May between the co-legislators - the European Parliament and the Council of the European Union – on the Digital Omnibus on AI. This provisional agreement must be now endorsed by the Council and the European Parliament before being submitted to a legal/linguistic revision with a view to the formal adoption of the legislative act by the co-legislators in the coming weeks.