The European Parliament has passed the Artificial Intelligence Act on 14th June 2023. The AI Act represents a set of regulations which aim to control the use of AI across the EU and promote safe, ethical usage of these systems.
DIGITALEUROPE is an organisation representing the digital technology industry across Europe, with members including major technology companies and national trade associations.
With the AI Act having the potential to impact Europe-based businesses, DIGITALEUROPE set out to find out just how much it could affect its members.
Savanta’s role in sandboxing the AI Act proposal
The primary objective of the AI Act is to ensure that AI systems respect fundamental rights and are under human supervision.
DIGITALEUROPE commissioned Savanta to conduct a pre-regulatory sandbox exercise consisting of a period of engagement and qualitative research amongst representatives of selected start-ups and SMEs across Europe.
Using our in-house qualitative expertise, we conducted nine in-depth interviews with senior stakeholders and delved into the expected impact of the proposed AI Act on their businesses and the tech industry more broadly.
The research provided multi-layered insight into both the challenges and advocacies of the AI Act across businesses. Whilst participants were genuinely committed to ethical AI usage and positive to the need for AI regulation, we found that its lack of clarity around key concepts and strict requirements could mean businesses will need to invest significant resources to ensure compliance. Such investments could pose challenges to smaller businesses that have limited budgets and resources, therefore reducing their ability to effectively innovate and compete in the wider market.
What exactly is sandboxing?
AI systems deal with vast amounts of data, and often with highly complex algorithms. Sometimes these algorithms can behave in unexpected ways, such as malfunctioning or glitching. Sandboxing is a way to mitigate the risk of these malfunctions, by isolating the AI system in a highly controlled environment.
This method aims to live test new policies, to understand the viability of these proposals. The point of a sandbox is to allow companies to “play” without regulations to collect evidence on the benefits and risks of innovative policy proposals. By doing so, companies can identify any problematic themes and recommend solutions to regulators. This approach has a number of synonyms, including policy probing or prototyping.
Why is it important?
While sandboxing is a new concept, we have seen an uptake of regulatory sandboxes in a number of sectors, such as financial services, tech and general innovation.
The key benefit of sandboxing is the ability to offer evidence on the benefits and risks associated with proposed policies. This allows regulators and companies to make informed decisions on necessary changes before new guidelines are rolled out, supporting innovation while mitigating risks.
Businesses included in sandboxes improve their understanding of new rules, which enables them to prepare for integration into their business plans and shape their products accordingly. For this reason, regulatory sandboxes are helpful for start-ups and SMEs which may find it difficult to deal with regulations due to having little influence, and financial constraints.
Most importantly, sandboxes allow regulators to test the viability of proposed rules, their impact on business and compliance strategies, as well as its wider effect on business landscape before committing to a larger (and costlier) deployment.
What is the impact of this project?
On a high level, this pre-regulatory sandbox exercise aimed to establish dialogue between members of the tech industry and EU lawmakers, bringing different perspectives to the table for the final stages of drafting of the AI Act.
The findings were captured in the ‘Sandboxing the AI Act’ report which was launched at DIGITALEUROPE’s annual Summer Summit. The report informed discussions over major themes that feed into Europe’s digital decade strategy.
This research highlighted the importance of multi-stakeholder engagement in policy testing. AI has become a global discussion – and there is a growing consensus that regulation is key – however, it is crucial that it is formulated correctly and fairly. By connecting and collaborating with stakeholders, who have a wealth of knowledge and experience within the field, better conclusions can be drawn, and fairer policies can be made.
Whether it is policy and communication development, testing and evaluation, or stakeholder research projects, Savanta is committed to providing insights and recommendations that will make an impact. Without research, there is no evidence to advocate for change.