European Commission white paper on artificial intelligence: our response

The need for a new regulatory framework

We agree that there is a pressing need for a common European AI regulatory framework to address the risk of fragmentation in the common market. Such fragmentation would be to the detriment of both citizens and businesses, inside and outside the EU. A common, predictable and legally certain regulatory framework is essential to driving innovation and adoption of AI.

We also see such a framework as an opportunity for the EU to be a global standard setter in AI – ensuring innovation and ethics go hand in hand. We welcome the EU's active engagement in global fora including the UN and OECD to promote AI that respects "fundamental rights, including human dignity, pluralism, inclusion, non-discrimination and protection of privacy and personal data" – European values which Scotland shares.

We commend the Commission's principle that "the new regulatory framework for AI should be effective to achieve its objectives while not being excessively prescriptive so that it could create a disproportionate burden, especially for SMEs." The Scottish Government's own Better Regulation agenda^[6] aims to reduce unnecessary burdens on business by ensuring all regulation is proportionate, consistent, accountable, transparent, and targeted only where needed.

To achieve this goal, we concur that "there is a need to examine whether current legislation is able to address the risks of AI and can be effectively enforced, whether adaptations of the legislation are needed, or whether new legislation is needed", in this order. We welcome the investigative work already carried out by the Commission on the safety and liability implications of AI, as published in the companion Report to this White Paper. We would encourage this work to be widened to a comprehensive review of all potentially relevant current legislation, to inform future regulatory mechanisms. 

We also agree that AI regulation must be both agile ("given how fast AI is evolving, the regulatory framework must leave room to cater for further developments") and pragmatic ("any changes should be limited to clearly identified problems for which feasible solutions exist"). To help achieve those objectives, we would recommend that the Commission consider the use of sector-specific regulatory sandboxes. These would need to be carefully designed, taking into account lessons learned from existing sandboxes, and adequately resourced – and would therefore benefit from collaboration between regulators in the EU. The evidence gathered from such controlled experimental approaches could be useful to assess the need for both horizontal and sectoral regulation of AI. Conversely, sandboxes could help regulators "enhance their AI skills in order to effectively and efficiently implement relevant rules", as proposed in the White Paper's section on Skills, and help businesses establish the processes required to develop safe and trustworthy AI.

More broadly, we share the Commission's working assumption that "the regulatory framework would apply to products and services relying on AI", that is, not to AI per se. Given the vast range of potential applications of AI, many of which "we can only begin to imagine", it seems prudent that new legislation required to address identified regulatory gaps be introduced progressively and from the bottom up, to avoid unforeseen negative consequences or obsolescence. Sectoral regulation, carefully tailored to each AI use case, built on the foundation of existing practice and expertise and further developed through the use of sandboxes where appropriate, could form the pillars upon which horizontal legislation would rest. This would allow the latter to focus entirely on addressing cross-cutting goals (such as upholding fundamental rights) and themes (such as data and decision-making) where these are not adequately covered by existing legislation.