Niamh Smyth introduces AI Bill to implement EU AI Act

I call on the Minister, Niamh Smith, to move the second reading motion as follows that the bill now be read a second time. I move. Thank you, Minister. You have 20 minutes, or their part of. Thank you, Ciann Comhairle. Ciann Comhairle, deputies, I move the regulation of Artificial Intelligence Bill 2026 now be read for a second time, and I'm very pleased to bring this bill before the House today. The purpose of this bill is to give full effect in Ireland to the EU Regulation 2024-1689 of the European Parliament and of the Council, and of the EU's Artificial Intelligence Act, which I will refer to throughout as the AI Act. The AI Act is designed to provide a high level of protection for people's health, their safety and fundamental rights, and to promote the adoption of human-centric, trustworthy AI. As with all EU regulations, the AI Act has a direct legal effect across all Member States. It aims to harmonise rules for AI systems across the EU, creating a single market and ensuring a consistent approach to risk, safety and fundamental rights. With those aims in mind, this bill is technically implementing that measure. It does not add to nor alter the obligations placed on regulated entities by the AI Act. The bill establishes essential domestic regulatory architecture for the supervision and enforcement of the AI Act in the state. A central feature of the bill is the establishing of the Office na hÉireann, the AI Office of Ireland, as an independent statutory body. Our new AI Office will be designated as the single point of contact for our citizens at home, the European Commission and other Member States. In this role, it will support consistent enforcement of the Act and facilitate cooperation between competent authorities and fundamental rights bodies. The Office will promote AI innovation and literacy and oversee the AI Regulatory Sandbox, which will foster innovation and support start-ups and SMEs in compliance with the AI Act. The bill also empowers our competent authorities and creates a robust administrative sanctions framework. This bill establishes a regulatory foundation on which to build trust in society for an evolution of AI that serves as a tool for people, with the ultimate aim of increasing human wellbeing. Artificial intelligence is one of the defining technologies of our era. Its applications now reach into virtually every corner of the economy and society, from healthcare, financial services, to the delivery of public services and much, much more. AI has brought remarkable benefits. It is accelerating scientific discovery, improving outcomes in medicines, expanding access to services and generating significant economic opportunity. Our national digital AI strategy, Digital Ireland, Connecting Our People, Securing Our Future sets out Ireland's ambition to be a global hub for applied AI innovation. But AI also carries real risk. Without appropriate oversight, it can embed and amplify discrimination, manipulate behaviour or exploit the vulnerabilities of people, posing threats to health, safety and fundamental rights of our citizens. The AI Act, which entered into force in August 2024, is the first ever legal framework designed to address these concerns, while simultaneously fostering innovation of responsible and ethical AI. The AI Act is a risk-based, horizontal regulation. It applies across all sectors of the economy, both private and public, with targeted exemptions for national defence, national security and purely personal use. Crucially, its measures are proportionate. Most AI systems will only give rise to minimal risk and can be marketed and used as subject to existing legislation without additional obligations under the AI Act. This ensures the EU and Ireland remain competitive for AI investment and innovation. The AI Act also takes into account specific interests and needs of SMEs in terms of the simplified obligations and tailored supports. There are five main pillars to the AI Act, which set out obligations across a number of categories and uses of AI systems and models. On prohibited practices, eight AI practices are prohibited due to the unacceptable risks they pose to European values and fundamental rights, including practices such as harmful manipulation, social scoring, real-time remote biometric identification. Given the rapid development of the technology, the AI Act also provides for the list of prohibited practices to be kept under regular review. Stringent requirements apply to high-risk AI systems used in eight areas set out in the AI Act, such as biometrics and critical infrastructure. Under transparency obligations, specific requirements apply to AI systems that interact with users, that generate synthetic content with risks of impersonation, or that enable deep fakes. Under the general-purpose AI pillar, providers of general-purpose AI models face obligations to mitigate systemic risks. Enforcement will be led by the European Commission with member-state cooperation. Under the enforcement pillar, member-states are required to designate competent authorities, to supervise and enforce the regulation of AI, and to provide for significant penalties with fines of up to £35 million, or 7% of global annual turnover, for the most serious infringements. I will now speak to the importance and urgency of the legislation. Ireland is a key position in the European regulatory landscape. Eight of the leading providers of the foundational AI models are established here on this island. These are the companies at the frontier of AI development and deployment globally. Maintaining the competitiveness and viability of Ireland's digital economy is an overarching national imperative, as outlined in the Programme for Government 2025, Securing Ireland's Future, and is the driver behind a number of key commitments in the Government's recently updated National AI and Digital Strategy. As Ireland assumes the presidency of the Council of the EU, meeting our obligations under the AI Act will further strengthen our position as an EU centre of excellence and a digital regulatory hub. It will also demonstrate our commitment to fostering responsibility, human-centric AI innovation, both domestically and across Europe. There is also a firm legal deadline. Implementation of prohibited AI practices posing unacceptable risks, as set out in Article 5 of the AI Act, takes effect as of 2 August 2026. Ireland must have the necessary national supervisory and enforcement infrastructure in place by that date. I want to acknowledge the considerable complexity and pace of work required to reach that point. The EU allowed a comparatively short window between the AI Act entering into force and its application. The complexity of the Bill, combined with the parallel development of the EU's Digital Omnibus Proposal, has required extraordinary efforts from my officials, from the Office of Parliamentary Counsel, and from the Office of the Attorney General. To them and to you, I am most grateful. As I have noted, and deputies, you will no doubt be aware, the EU's Digital Omnibus on AI, a set of targeted amendments to the AI Act, was provisionally agreed in May 2026 and approved by the European Parliament in June. Ireland supports the agreed Omnibus, which helps increase legal certainty and reduce administrative burden. Importantly, the Omnibus extends certain implementation timelines, including for the transparency obligations and high-risk AI systems requirements. However, the Omnibus does not extend the timeline for enforcement of prohibited AI practices under Article 5. The 2 August 2026 deadline remains firm, and this Bill addresses it directly. Any amendments to domestic legislation that may be required following formal adoption of the Digital Omnibus on AI will be addressed as part of Phase 2 of this work in the autumn. The second phase will also provide an opportunity to further enhance AI regulation in accordance with the AI Act, if required. So the overview of the Bill. Having set out the context and the urgency of the Bill, I will now outline its provisions. The Bill is structured in ten parts, with 139 sections and four schedules. Its overarching policy objective is to ensure that the AI Act can be comprehensively and effectively implemented here in Ireland. Part 1 of the Bill, comprising Sections 1 to 5, provides for the title and commencement, key definitions, regulation-making powers, services of documents and expenses to provide the cost of administrating the Act, and is funded with the appropriate ministerial approval. Part 2 is set out over Sections 6 to 40, and provides for the establishment of IFIC-IS and its governance structure, including appointment of its Board, CEO and staff, its accountability, its funding and its functions. Sections 12 to 23 are specifically concerned with matters relating to Board governance, with provisions on Board establishment and membership, appointments process, terms of appointment, membership of either House of the Oireachtas, remuneration and conflict of interest and others. Sections 24 to 33 relate to the Chief Executive Officer role, and provide for the matters relating to the CEO's appointment, resignation, accountability and functions. Sections 34 and 36 relate to the staff of IFIC-IS, na h-earn, and confidentiality. The appointment of staff is subject to the Public Services Management Act 2004 and the Civil Service Regulation Act 1956-2005. Section 37 to 40 provides for funding and accountability, including the strategy statements, accounts and annual report. Section 3, sorry, Part 3, comprises Sections 41 to 52, provides for the practical operation infrastructure underpinning enforcement of the AI Act in Ireland. Section 42 designates the AI office as the national single point of contact in relation to the public and to other counterparts at member states and union levels, and will help increase organisational efficiencies as required by the AI Act. Section 43 provides that the AI office will establish and maintain a national AI register, recording all incidences of prohibited AI practices, serious instances involving high-risk AI systems, and other required notifications. Privacy and public accountability are embedded from the outset. Section 46 provides for the establishment of the AI regulatory sandboxes under the supervision of the AI office. SMEs and start-ups are to receive priority access. Sections 48 to 50 make provision for real-world testing of high-risk AI systems in controlled conditions, subject to appropriate safeguards, oversights and rights of review. Market surveillance authorities are a key cohort of the national competent authorities required for the implementation of the AI Act, and Part 4 sets out, as Sections 54 to 62, the general obligations and powers of market surveillance authorities, including the power to receive complaints from the public. Section 59 enables fundamental rights bodies to request the testing of high-risk AI systems, where infringements of fundamental rights are suspected. Sections 63 to 64 are concerned with cooperation agreements of relevant market surveillance authorities. Sections 65 to 66 set out data protection matters for relevant competent authorities. Section 5 provides market surveillance authorities with a comprehensive, structured and proportionate enforcement toolkit, progressing from cooperative compliance measures through to cohesive interventions. Authorised officers appointed by each market surveillance authority under Sections 68 to 70 will have powers to enter premises, to inspect relevant AI products and records, to require the production of information and to take samples. Section 71 provides for the contravention notices may be issued requiring an operator to remedy a breach. Section 72 provides the prohibition notice may be issued to stop an AI system from being placed on the market or put into service where it poses a risk or breaches the AI Act. Sections 73 and 74 provide powers for seizure, disposal and forfeiture of unsafe AI products and under Section 75, in the most serious cases, for the removal of online content to eliminate serious risk. Section 77 provides rights of appeal to the district court throughout. Part 6 establishes at Sections 78 to 120 the independent adjudication system for the imposition of administrative fines for breaches of the AI Act. Sections 79 to 86 provides that where market surveillance authorities believe an infringement has occurred and an authorised office has conducted an investigation, the matter may, with the authority's consent, be referred to an independent adjudicator. The administrative fines set out in Section 105 are substantial and fully aligned with the fines set out under the AI Act. For public bodies, fines are capped at one million, in line with data protection legislation. Sections 114 provides that no administrative fine takes effect without confirmation by the High Court, which is a key procedural safeguard. Sections 107 to 109 provides for the rights of appeal. Sections 116 to 117 provides for the nomination of adjudicators by the AI Office and appointment by the Minister for Enterprise, Tourism and Employment. Sections 118 expressly provides for the adjudicators' independence in the performance of their functions. Part 7 of the Bill concerns penalties and miscellaneous provisions at Sections 121 to 127 sets out criminal penalties for obstruction of authorised officers, failure to comply with enforcement notices and unauthorised disclosure of confidential information. Part 8 at Sections 128 to 133 amends the Central Bank Act 1942 to equip the Central Bank with the investigatory and sanctioning tools necessary for its role as a market surveillance authority under the AI Act. Part 9 at Sections 134 and 137 amends the Competition and Consumer Protection Act 2014 to equip the Competition and Consumer Protection Commission with the investigatory and sanctioning tools necessary for its role as a market surveillance authority under the AI Act. Part 10 at Sections 138 and 139 contains miscellaneous amendments including consequential amendments to Communications Regulations Act 2002 and the Freedom of Information Act 2014. Additionally, there are four schedules governing oral hearing procedures, the assignment of adjudicators and the register of adjudicators. In conclusion, Can Córla, establishing a credible and coherent AI enforcement regime in Ireland, underscores our commitment to responsible AI development and is a core component of our international competitiveness offering. In this respect, the Regulation of Artificial Intelligence Bill 2026 matters to how AI will impact our society, our economy and international standing and, of course, our future. It puts in place the enforcement structures that will protect our people in Ireland from AI systems that pose unacceptable risks to their health, to their safety and to their fundamental rights. It provides a regulatory clarity and helps build the public trust that businesses, both indigenous and multinational, need to invest, to operate and to grow with confidence. As we assume the Presidency of the Council of the EU, meeting our obligations under the AI Act demonstrate our commitment to responsible digital governance and to being a trusted partner in the EU. The implementing measures enabled by the legislation, including the establishment of a new AI office in Ireland, will provide an enduring foundation, strengthen our position as the EU centre of excellence and digital regulatory hub and demonstrate our commitment to fostering and promoting responsible AI innovation and adopting both in Ireland and in Europe. It is imperative that the State has a comprehensive and robust legal basis for the full and effective implementation of the AI Act. I am confident that this Bill achieves that objective in a balanced, proportionate and technically sound manner.