The technological development of artificial intelligence (AI) is progressing rapidly, with investment in AI-based solutions dramatically increasing worldwide. Europe is no exception. Given the enormous impact these technologies have on citizens and their fundamental rights, regulatory practices need to be put in place. The European Commission has a clear strategy to promote the digital transformation and – in an effort to protect citizens’ fundamental rights – the digital space and AI are increasingly regulated. To this end, in April 2021, the Commission proposed a regulatory framework on AI (AI Act). To ensure the fairness and reliability of all high-risk AI systems that are deployed in Europe, the AI Act will require AI systems to undergo conformity and conformity assessments, which will be costly for businesses, especially SMEs. SMEs are at the forefront of innovation in Europe, but they are easy to remedy by high compliance costs, risking an unsustainable burden.
The main objective of the study was to estimate the costs SMEs would face to ensure compliance with the requirements of the AI Act. To this end, we have considered the following four scenarios:
- Scenario 1 adapted the standard IARR cost model – which calculates the total costs as the sum of compliance and compliance costs – to the modelled SME, dividing the total market value of AI by the value of one unit of AI, then multiplying it with compliance costs.
- While following the standard cost model approach, Scenario 2 calculated compliance costs based on the value of C &Dinstead of the total market value of AI.
- Scenario 3 provides an alternative approach whereby total costs are not calculated as a percentage of the value of AI, but rather consist of software development costs and costs for compliance activities related to C & D.
- A fourth scenario (scenario 3b) is based on scenario 3 assumptions and also includes the cost savings for SMEs resulting from the involvement of European Digital Innovation Hubs (EDIH) and Testing and Experimentation Facilities (TEFs).
The results of the analysis show that if compliance costs are calculated as in scenario 3 (EUR 301,200), they are more than 10 times lower than the costs calculated as a direct proportion of the market value of AI (around EUR 4 million in scenario 1). When EDIH and TEF are included in the analysis (scenario 3b), the costs are expected to decrease further (EUR 229,444), as are the FTEs needed to comply with the requirements.
As demonstrated in this document, the EDIH and the TEF will play a significant role in mitigating compliance costs for SMEs when adhering to the requirements of the AI Act. Timely involvement of EDIH and TEF – through the development of common technological, legal and management services – is recommended to activate economies of scale and generate cost savings for SMEs when complying with the new requirements. Further research on the benefits that SMEs in the EDIH and TEF could benefit would help to strengthen the case for AI uptake.