The SMARD-GOV research project investigates how large language models (LLMs) can support decision-making processes in public administration in a secure, privacy-preserving, and legally compliant manner. The project is funded by the German Federal Ministry of Research, Technology and Space (BMFTR) and is jointly carried out by Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), the University of Münster, and PROSOZ Herten GmbH.

Over the past months, the project has generated new insights and presented its research findings at several academic and professional events.

As part of the Computers, Privacy and Data Protection (CPDP) conference in Brussels, the University of Münster organized an interdisciplinary workshop on data accuracy in LLMs. Together with journalist Martin Bernklau, legal expert Constantin Lessmann (Friedrich-Alexander-Universität Erlangen-Nürnberg) and computer scientist Kristina Magnussen (University of Münster) discussed the risks of inaccurate AI-generated statements about individuals, the legal requirements of the General Data Protection Regulation (GDPR), and current technical challenges. The workshop highlighted the continued need for research to develop trustworthy, privacy-preserving AI systems.

At the end of June, Johannes von Ivernois (University of Münster) participated in a hackathon organized by the German Federal Ministry for Digital and State Modernization (BMDS) to further develop the project "Faster Planning and Implementation through AI (SPARK)" in Berlin. As SPARK shares several thematic overlaps with SMARD-GOV, the event provided valuable opportunities to gain new insights for the project's research while contributing SMARD-GOV's findings to the further development of SPARK.

At the MEMO Conference at the University of Münster, Prof. Tobias Brandt presented initial findings from interviews and a pilot experimental study. The results were discussed with experts from public administration, fostering an exchange on current challenges and opportunities related to AI-supported administrative processes.

In addition, the SMARD-GOV team organized two expert workshops addressing the question: "How can LLMs be integrated into administrative procedures in a legally compliant way?". The workshops were aimed at stakeholders from the municipal sector and provided insights into initial project results while facilitating discussions on legal and technical challenges.

Research Partners Wanted

SMARD-GOV is currently looking for research partners from public administration to support the further development and validation of reliable and privacy-preserving AI systems. Real-world application data is needed to evaluate the developed approaches under practical conditions. Partners will gain early insights into project results while contributing to the development of privacy-preserving and legally compliant AI solutions for public administration.

Interested in collaborating? Contact the SMARD-GOV project team: Constantin Lessmann (constantin.lessmann@fau.de).