Ireland’s long journey to ‘trustworthy AI’
Ahead of recent agreement on a new European AI Act, the Government has approved guidance for the use of AI in the public sector. However, the journey to achieve robust and effective regulation has faced significant challenges along the way.
Statements obtained from government departments by eolas Magazine outlined how there has been no single, unified approach taken by government departments for the use of artificial intelligence (AI) in the public sector.
Therefore, the Government has instructed that departments and public bodies should comply with seven requirements for ethical AI that have been developed by the European Commission’s High Level Expert Group on AI in its Ethical Guidelines for Trustworthy AI document.
In the document, the Government outlines the seven overriding principles which will shape how AI is used in the public sector. These are:
- human agency and oversight;
- technical robustness and safety;
- privacy and data governance;
- transparency;
- diversity, non-discrimination and fairness;
- societal and environmental well-being; and
- accountability.
The guidelines, which were shaped by work carried out by the High Level Expert Group on AI, state that public sector AI systems may be developed either in house or procured from a third vendor.
In either case, however, the guidelines clarify that there is a high level of responsibility for public sector bodies and departments to ensure that AI is trustworthy and ethical, and that this responsibility rests with the system user rather than the developer, thereby protecting private sector providers and developers. This point is underlined by the directive that this responsibility “cannot be delegated”.
This point is further emphasised by the stipulation that ethics considerations must be built into requirement specifications of systems, and that specific test cases must be written and signed off by senior management of the public sector organisation.
The guidelines further outline that subject matter experts should be part of the design process for an AI system and have “an important role to play in ensuring that all service users that will be affected by the system are treated fairly”.
Challenges along the journey
Regulating AI remains a difficult challenge for decision-makers, as it is a sector for which a one-size-fits-all approach is not applicable due to the different needs of the various strands of the public sector.
Speaking to eolas Magazine in February 2024, Government Chief Information Officer Barry Lowry said that he was “very pleased to note the progress made with the EU AI Act”.
“I think it has evolved well to now provide a very good balance between meaningful AI use and risk and privacy management. I look forward to its enactment,” Lowry said
As the EU AI Act was being drafted, Lowry had warned in a previous interview with eolas Magazine (issue 60) that “there is a risk that misunderstood attempts to regulate AI are viewing AI as an entity, but it is not about AI as an entity, it is about how you use it”.
Lowry had further warned of the potential risks of introducing legislation “without a real understanding of what it is that is being regulated” and “without talking about shared risk”.
Development of AI legislation has faced further challenges amid opposition from government TDs to the development of facial recognition technology (FRT), as some decision-makers have suggested that this may involve invasion of citizens’ rights to privacy.
Maintaining trust
In an effort to acknowledge these points of concern raised by various stakeholders around data privacy, the government stipulates that GDPR requirement will be enforced when data is being used in AI models.
Under GDPR legislation, permission must be sought to use personally identifiable information, which includes facial images and voice.
“Data should not be used in AI models in a way that breaches intellectual property rights. Where the system is procured from a third-party vendor, they must confirm that their data is GDPR compliant and does not breach the intellectual property rights of others,” the guidelines state.
Speaking following the publication of the new guidelines for AI in the public sector, Minister for Public Expenditure, NDP Delivery and Reform Paschal Donohoe TD said:
“AI presents significant opportunities to improve the delivery of public services and to assist in tackling many of the complex challenges we face. Our government and public service must not shy away from the opportunities presented by AI because of the potential risks involved.
“The guidance issued… will help public servants avail of those opportunities in a way that is ethical and maintains trust. AI is changing our world significantly and it is important we face that change directly and adapt quickly as required to both protect and deliver for the people we serve.”
