Europe on the brink of setting a decisive course for artificial intelligence

Berlin, 5 December 2023 - Tomorrow’s political trilogue negotiations on the AI Act could be conclusive. The European co-legislators will meet to reach an agreement on the AI Act, which aims to regulate the development and application of artificial intelligence in Europe. Bitkom President Dr Ralf Wintergerst explains: 

“Europe is about to set the course for artificial intelligence that will have enormous significance for society and the economy. The EU AI Act will comprehensively define the legal framework for the application and development of AI technologies in Europe and will decide how AI can be used in Europe in the future - but also how and whether companies can develop AI in Europe. The AI Act can help to create the necessary trust in and acceptance of AI systems through ethical standards and safety requirements. The AI Act must now deliver on its promise to foster innovation. The rules must be flexible and practical so as not to stifle the dynamic development of AI technology before it has really begun.

The remaining key question is whether and how the AI Act will deal with AI models, i.e. foundation models or general-purpose AI models. Their strict and rigid regulation would be problematic for two reasons: First, providers cannot anticipate the multiple uses of their models, making effective risk assessment and mitigation impossible. Second, the pace of technological development, particularly at the modelling level, is so rapid that rigid rules in the AI Act would quickly become obsolete. The better way forward is the compromise proposed by Germany, France and Italy. This provides for the necessary requirements to be laid down in binding self-regulation in cooperation with the model developers. The advantage of this approach is that it is both feasible and can be dynamically adapted to technical developments. To be clear, binding self-regulation does not mean that there are no rules. Rather, it means that the stakeholders involved - companies, research institutions, developers, associations and civil society – will work together to establish the necessary requirements for AI models in a flexible and practical manner. These can be very specific and include, for example, technical documentation, ethical principles or safety standards, but without running the risk of creating unnecessary hurdles through overly rigid rules that disconnect or exclude Europe from the development and use of AI.”