Many companies are currently holding their breath ahead of Parliament’s upcoming vote on the AI Act. And maybe many more of us should be doing the same? A great deal is at stake for how we will be able to take advantage of AI in Europe in the future.
From a company perspective, it is about competitiveness, market share and job opportunities. The AI proposal has been under discussion in Parliament for months and initially proposed many relevant and important improvements to the European Commission’s proposal of April 2021. In recent weeks, however, these changes have now become so extensive and wide-ranging that they are difficult to interpret and their consequences challenging to predict. Some companies are insisting that the administrative burdens for using so-called ‘high-risk’ and ‘generative’ AI will lead to delays of up to two years before a new product or service can be put on the market. This does not work in practice. By then, Europe will have lost ground in the rapid innovation and use of beneficial new technologies. This would negatively affect the entire European data economy.
In addition to the AI Act, the Data Act is also being negotiated. The final version of these two Regulations will affect the business models and business opportunities extensively. As it currently stands, parliamentarians cannot possibly have had sufficient time to complete their analysis satisfactorily. More time and greater (AI) knowledge is needed in order for these Regulations to turn out correctly and be capable of benefitting both business and wider society.
If we are to achieve the best possible outcome, it would be worth slowing down the process and taking time to rethink. It will be difficult to include all the different definitions and techniques, particularly as AI research is advancing so rapidly. New techniques are emerging before existing ones have even been legislated for. In the best of worlds - as the proposal currently stands - we should skip over the various AI-techniques and approaches definitions. Instead, legislators should focus more on the uses and outcomes of AI systems based on a risk-based approach, general principles for trustworthy AI (Article 4a), prohibited practices (Article 5), classification rules for high-risk AI systems (Article 6), transparency obligations for AI systems, (Article 52) and high-risk areas (ANNEX III). This approach would be much more achievable, understandable, and predictable.
Many of the AI Act’s other approximately 80 articles impose a significant bureaucratisation and administrative burden for private and public sectors. Those articles would need to undergo a new impact assessment whether they are proportionate and proportionately contribute to trustworthy AI in addition to any existing legislation.
If legislators take their time, EU could achieve a much more precise, robust and technology-neutral legislation in line with the Better Regulation agenda. This would likely strengthen the desired AI innovation and use in Europe.
Objectives of the Better Regulation agenda
· Ensure EU policymaking is based on evidence
· Making EU laws simpler and better, and avoiding unnecessary burdens
· Involving citizens, businesses and stakeholders in the decision-making process
Ethical guidelines for Trustworthy AIAI Act