Training of machine learning (ML) models is practically very often done with data that has a personal reference. The resulting challenges for compliance with the GDPR are extensively addressed in the literature. As a result, training is often made more difficult and can only be designed to be legally compliant to a limited extent. This conflict is addressed by the legal acts of the so-called EU data law, some of which have already been passed and some of which are still in the legislative process.

In addition to the planned AI Regulation, the provisions of the Data Governance Act, the Data Act as well as the regulations for specific data spaces – such as the European space for health data – could also significantly influence the training of ML models in the future.

On the one hand, the aim is to create improved access to data; on the other hand, specific regulations for the development of AI systems may enable the processing of personal data in compliance with the GDPR.

This presentation will provide an overview of how data law will affect the training of ML models in the future.