Bitkom on the vote on the AI Act  

• EU Parliament decides on the AI Act today - then the trilogue negotiations begin
• Berg: "The AI Act must not become an AI prevention act".

Berlin, 14 June 2023 - Today, the European Parliament votes on the AI Act, which aims to create EU-wide rules on the use of artificial intelligence. This clears the way for trilogue negotiations, in which the European Parliament, European Commission and European Council negotiate the final version of the regulation. Bitkom President Achim Berg states:

"Europe is at a crossroads when it comes to artificial intelligence. Now it will be decided whether the AI Act will hinder AI innovations in Europe and make their use more difficult or whether it will open up a perspectives and opportunities for AI in Europe. With today's vote in the European Parliament, the discussion on the AI Act is not over, it is entering its decisive phase.

The task now must be to correct the weaknesses of the proposal on the home straight in the legislative process. The underlying risk-based approach is good and correct. Only for the use of AI in the so-called high-risk area should strict requirements apply to ensure that AI systems are safe and trustworthy. But if almost every second AI application were to be assigned to this high-risk area, this would lead the approach ad absurdum. It is important to strictly regulate only those applications that can actually pose massive dangers and risks in practice. Additionally, the AI Act must clearly demarcate which requirements apply to whom in order to avoid additional legal uncertainty. A narrow and legally secure definition of AI itself is also central; the AI Act must not become a Software Act that regulates all statistical systems or even almost all software.

The use of AI is already covered by a multitude of regulations and laws, from data protection to consumer law. Therefore, for example, questions of data protection, copyright law as well as rules for platforms and social networks or questions of sustainability do not belong in the focus of the AI Act.  It should focus on defining requirements for market access of safe and trustworthy AI systems and products in which AI is integrated. As originally planned, the EU should continue to follow exclusively the guiding principle of the "New Legislative Framework", which is already known to citizens at least indirectly through the CE mark.  The latest developments in the field of generative AI, such as ChatGPT, have also shown that technological development is highly dynamic. In the first regulatory proposal presented by the EU Commission in 2021, generative AI still played a minor role. It is important to ensure clarity and legal certainty for generative AI in the AI Act, with specifications that also work in practice.

In principle, the EU and the member states should not only create abstract rules, but also the right framework conditions for the successful development of AI and for a practical implementation of regulation. To implement the AI Act, we need supervisory and market monitoring structures as well as sandboxes as experimental spaces that help companies to implement regulation in a practical and pragmatic way. We must not put obstacles in the way of those who want to advance AI in Europe. The AI Act must not become an AI prevention act, it must become an AI acceleration act. The goal must be for people and businesses in the EU to experience and fully benefit from the advantages of AI."