|
Linnea Decker

SMARD-GOV: Data Protection Compliant Large Language Models (LLMs) for Public Authorities

With the research project SMARD-GOV, Friedrich-Alexander University Erlangen–Nuremberg (FAU), the University of Münster, and the software and consulting company Prosoz Herten GmbH are investigating the use of data protection compliant Large Language Models (LLMs) in public authorities. The aim is to develop secure and legally compliant AI-based decision-making systems, exemplified by applications in the area of building permit procedures. The project combines technical, legal, and social science research and takes into account the specific requirements of public administration.

Prof. Paulina Jo Pesch from FAU coordinates the project and, together with her team, analyzes the legal framework for the use of LLMs in data protection and AI regulation. Prosoz, represented by CDO and AI expert Christian Rupp and his AI team, contributes its practical experience in software development for municipalities and supports the project through its professional network. At the University of Münster, the IT Security research group led by Prof. Rainer Böhme conducts training sessions and tests to evaluate the effectiveness of data protection measures in LLMs. In addition, the Chair of Digital Innovation and the Public Sector, headed by Prof. Tobias Brandt, is conducting empirical research into the use of LLMs in public authorities.

The project is funded by the Federal Ministry of Research, Technology and Space (BMFTR) from 2025 to 2027 and will make an important contribution to the privacy-compliant use of modern AI technologies in the public sector.