I thank colleagues very much for their amendments. I will try to go through each and every one. The purpose of section 9(1)(c) is to provide a broad statement of the office's function to promote and foster AI innovation and literacy. The Bill is intended to give effect to the EU AI Act and in doing so it is important that the language used in the Bill remains consistent with the terminology and objectives of the underlying regulation.
Article 1 of the EU AI Act states that its purpose "is to improve the functioning of the internal market and promote the uptake of human-centric and trustworthy artificial intelligence (AI), while ensuring a high level of protection of health, safety, fundamental rights ... and supporting innovation."
The regulation, therefore, already embodies the principle that Deputy Gibney's amendment seeks to advance.
In addition, neither "ethical" nor "responsible" is defined in the AI Act or in the Bill. Introducing these terms in a single provision without definition and without corresponding usage elsewhere in the legislative framework could create uncertainty as to whether a different or additional standard is intended to apply to the office's functions. I consider that the existing reference to promoting and fostering AI innovation and literacy is sufficiently broad and that the wider framework of the AI Act already ensures that innovation is pursued in a manner that is human-centric, trustworthy and protective of fundamental rights. Accordingly, I do not consider the amendment necessary and I do not propose to accept it.
While I appreciate the intention behind the amendment, these concepts are already inherent in the AI Act framework, and the current drafting is clear and consistent with the language of the underlying regulation. Recital 8 of the AI Act states the regulation supports the objective of promoting the European-centric approach to AI and making the Union a global leader in the development of secure, trustworthy and ethical AI, reflecting the conclusions adopted by the European Council in October 2020. The same recital further notes that the regulation ensures the protection of ethical principles, reflecting recommendations made by the European Parliament in its October 2020 resolution on the ethical aspects of artificial intelligence, robotics and related technologies.
Moving to amendment No. 3, the Bill already considers a number of the provisions designed to ensure transparency and accountability in the performance of the office's function, including requirements to prepare a strategy statement and an annual report, both of which must be published. In addition, section 9(4) provides that the office shall have all such powers as are necessary or expedient for the performance of its functions.
Where research, analysis or the publication of reports is required to support the exercise of these functions, the office would already be able to undertake such activities. The proposed amendment would introduce a broad stand-alone function to prepare and publish reports on research undertaken, sponsored, commissioned or assisted by the office. It is not clear what additional legal effect this would have, particularly as the office would already be capable of publishing research or reports in support of the performance of its statutory functions, including but not limited to functions under section 9(1)(c) and 9(1)(d) to promote AI innovation and literacy and to enhance public awareness. I am confident the Bill already provides an appropriate framework and that the amendment is not necessary.
The Government is conscious that the office is being established as a new body and its functions may evolve as the AI regulatory framework becomes more established and operational experience is gained. For this reason, the Bill seeks to provide an appropriate foundation while avoiding unnecessary prescription in the activities the office may already undertake in support of existing functions. I am nonetheless sympathetic to the objective underlining the amendment. The role of research, evidence-gathering and public reporting in supporting effective AI governance will be kept under constant review by the office as it develops and as its work programme matures. Should there be a demonstrated need to place such activities on a more explicit statutory footing in future, this can be considered in light of operational experience. Accordingly, while I appreciate the intention behind the amendment and the value of research in informing AI policy and regulation, I am satisfied the Bill already provides the office with sufficient flexibility in this regard and I do not propose to accept the amendment at this time.
On amendment No. 4, section 9(1)(c) provides that the office shall promote and foster AI innovation and literacy and section 9(1)(d) provides that it shall enhance public awareness of the use of AI systems and the rights and obligations under the artificial intelligence regulation. These functions complement the range of supports that are provided for in the AI Act at EU level, including guidelines, codes of practice and a service desk, which are intended to support and reinforce consistent implementation of harmonised rules across the EU. I am satisfied that these provisions already capture the proposed paragraphs (h), (i) and (j), and reinstating them in granular subparagraphs would result in legislative duplication without adding substantive legal effect.
The Bill, through section 46, provides for the establishment of one or more AI regulatory sandboxes in accordance with Article 57 of the EU AI Act, with priority access to SMEs and start-ups free of charge, as mandated by Article 62 of the regulation. The sandbox function is the primary structural vehicle through which the office will support SME compliance, provide guidance and technical assistance and reduce regulatory barriers to AI adoption. The proposed paragraph (k) duplicates this as it is already embedded in the sandbox and real-world testing framework provided for in Chapter 3 of Part 3. Section 9(1)(f) already requires that the office is to provide advice to the Minister on any matter relating to its functions. The proposed paragraph (j) requiring the office to provide advice and technical assistance to the Minister, competent authorities, businesses and organisations is repetition of the existing function. Section 10 confers on the Minister power to assign additional functions to the office by ministerial order following consultation with the office and relevant Ministers. This ensures that the office's functions can evolve responsibly to align with wider national policy priorities.
While I understand the sentiment behind amendment No. 5, I am satisfied the protections are already guaranteed under the AI Act, which has direct effect in Ireland. Asserting them as part of the AI office would not strengthen them but merely duplicate what is already binding in law. The Act completely prohibits AI systems that infer employees' emotions in the workplace or educational institutions. It also bans biometric categorisations that infer sensitive traits. The AI Act imposes a binding obligation on employers as deployers of high-risk AI systems to inform workers' representatives and affected workers before putting such systems into use in the workplace. Similarly, AI-driven decisions on employment status must have meaningful human oversight. Workers have the right to demand human intervention, obtain an explanation for a decision and contest adverse outcomes.
The AI Act is very clear on its interplay with other EU and national legislation. It explicitly calls out that the harmonised rules laid down in the AI Act should apply across sectors and in line with the new legislative framework, they should be without prejudice to existing Union law, in particular on data protection, consumer protection, fundamental rights, employment, protection of workers and product safety, to which the AI Act is complementary. Recital 9 of the AI states that, in the context of employment and the protection of workers, this regulation should not affect Union law on social policy and national labour law. Further, it states the regulation should not affect the exercise of fundamental rights as recognised by the member states and at Union level. Worker reskilling and upskilling in the context of AI adoption is already a matter of active Government policy, addressed through the national AI strategy, SOLAS, Skillnet Ireland, the National Training Fund and the further and higher education system in its broad mandate. The office will complement this work and work in co-operation with the relevant State actors active in this area.
The point made by Deputy Ó Murchú is correct and it emphasises the cross-party and horizontal impact the Government is having on AI. The meeting he referred to is the round-table conversation the Minister, Deputy Lawless, had on enterprise, with many stakeholders around the table. It was specifically on the gaps we need to fill and whether they are widening or narrowing. Admittedly, there is a widening between SMEs and the bigger companies which have all of these systems in place. We were rightly taking a look at how we can address this with Skillnet Ireland, our ETBs, SOLAS and many other training agencies.
To respond to the point made by Deputy Lawlor, the Bill seeks to provide an appropriate foundation while avoiding unnecessary measures. I am satisfied the AI Act already provides appropriate safeguards in the workplace and I do not propose to accept the amendment at this time.
Turning to amendment No. 6, I agree that it is important that the emerging issues in AI regulation are identified early and that the regulators work together and effectively in responding to new challenges. The office is being established primarily to support the implementation and co-ordination of the AI Act in the State. Its functions include co-ordinating competent authorities, facilitating co-operation and information sharing, supporting enforcement activities and advising the Minister on matters relating to its functions. These functions already place the office in a strong position to identify emerging issues and share information across the regulatory system. The amendment would go further by assigning the office a potentially broad and undefined role in "addressing [areas] where regulation may not be covered by existing authorities". It is unclear what legal or operational action that would entail and how it would interact with the responsibilities of the Oireachtas, the Government, Ministers and sectoral regulators.
The creation of new regulatory frameworks, the allocation of regulatory responsibilities and the identification of legislative gaps are primarily matters for the Government and the Oireachtas. The role of the office is to support effective implementation of the existing framework established by the AI Act and this Bill. That said, I acknowledge the concern underlined in the amendment. As the AI ecosystem evolves, the office's practical experience and engagement with competent authorities will place it in a valuable position to identify emerging regulatory issues and advise the Government accordingly. Indeed, section 9(1)(f) empowers the office to provide advice to the Minister on matters relating to its functions.
Amendment No. 7 is provided for under the EU AI Act, which introduces the provisions concerning energy consumption and transparency. Providers of the general purpose AI models are required to document the known or estimated energy consumption of their models as part of their technical documentation obligations under annex XI of the AI Act. To support this, the European Commission is currently undertaking work to develop a robust and practical framework for the measuring of AI energy consumption and efficiency. Moreover, the Bill requires the office to have regard to the "policies and objectives of the Government and any Minister of the Government insofar as the policies and objectives may affect or relate to the functions of the Office". This provision is broad and captures wide Government policies, including climate.
The AI Act is clear and strong in its emphasis on its application in accordance with the values of the Union enshrined in the charter, facilitating the protection of natural persons, undertakings, democracy, the rule of law and environmental protection in the uptake of trustworthy AI. The Government is conscious that the functions of the office may evolve over time but it is important to avoid unnecessary duplication in relation to its function. The amendment is therefore unnecessary, and I will not be in a position to accept it.
While I understand the sentiment behind amendments Nos. 8 and 9, I am satisfied that the section, within its broad mandate, places obligations on the office to promote AI innovation, adoption and competitiveness. The Bill also requires that the AI office is established to supervise AI regulatory sandboxes in accordance with Article 57 of the EU AI Act. The sandbox framework is specifically designed to provide a structured environment for innovation and testing and development, practically benefiting SMEs and start-ups that may lack the resources of larger enterprises to navigate compliance independently. The EU AI Act's sandbox provision was deliberately designed with SMEs in mind and the Bill faithfully implements that framework.
In relation to the amendment regarding annual reports, the Bill provides a comprehensive accountability framework that achieves this purpose. Section 38 requires the office to prepare a strategy statement every three years setting out key objectives, outputs and strategies, including a review of previous performance. Section 40 requires that the office submit an annual report on the performance of its functions in such form and including such information as the office considers appropriate or as the Minister directs. I am satisfied that the proposed objectives of the amendment are captured within the core functions of the office. Reasserting them would merely restate what is there, maybe in more elaborate terms. Accordingly, I do not accept the amendments.
Amendment No. 10 would require the office to take such measures as it considers necessary to advance the objectives specified in section 9(1) and to report on those measures in each annual report under section 40. Section 9(1) already confers on the office a comprehensive range of functions directed at precisely these objectives while section 9(4) provides that, "The Office shall have all such powers as are necessary or expedient for the performance of its functions", which already encompass the taking of measures that the proposed section 9(6) would require. Imposing a separate, open-ended obligation to take such measures as is considered necessary to advance those same objectives offers nothing of substance while potentially creating interpretive ambiguity as to the scope of the office's existing powers and duties. As regards the reporting obligation proposed in section 9(7), section 40 already requires the office to prepare an annual report on the performance of its functions, which inherently encompasses reporting on the measures taken in respect of the section 9(1) objectives. A specific and additional reporting requirement is therefore redundant.
I am conscious that the office must be afforded appropriate flexibility and independence in determining how best to advance its statutory objectives, a point expressly recognised in section 9(5). The prescribing of further layers of mandatory consultation and reporting obligations of the kind proposed risks imposing an administrative burden disproportionate to any tangible benefit. Accordingly, I oppose the amendment.
