Schalast | The draft AI Regulation
The use of artificial intelligence (AI) is currently experiencing rapid growth in importance, although the opportunities offered by the new technology are naturally also associated with legal risks. The draft AI regulation published by the European Commission on 21 April 2021 represents a key step towards a regulatory regime for AI.
The purpose of the AI Regulation is to create uniform standards for the use of AI for a new technology that is not yet specifically regulated. At the same time, the aim is to protect the fundamental rights concerned from the risks inherent in the technology. This is not only the first comprehensive regulatory approach in Europe, but worldwide.
As a regulation, the AI Regulation leads to a fully harmonised legal situation within the EU with a uniform level of protection. The draft regulation provides for a risk-based approach to address the risks of AI. A distinction is made between prohibited AI practices (Art. 5), high-risk AI systems (Art. 6), low-risk AI systems (Art. 52) and minimal-risk AI systems (Art. 69).
Depending on the form of AI, there are different requirements for the AI system. This risk-based approach should enable a proportionate and graduated response to the respective risk potential.
As is usual with such "major pieces of legislation", the drafts are the subject of intense debate. In this respect, the focus is on the broad scope of application. It is true that Art. 2 (2) of the AI Regulation may exclude some sectors - particularly transport and mobility. Nevertheless, it is important for companies, manufacturers and providers to note that the legal acts of the exempted sectors are to be adapted. This is because the requirements from the AI Regulation should also apply there and the special features of the respective sector, for example in the form of a type approval procedure or a market surveillance authority, should remain in place.
In addition, the AI Regulation not only stipulates obligations from the time of placing on the market, but also regulates obligations for developers. Numerous requirements for training, validation and test data sets are envisaged in order to avoid distortions in the data itself.
The final version of the AI Regulation is eagerly awaited.