Robotic systems have great potential to improve medical care, but their integration by non-technical staff remains a challenge. LLMs open up new possibilities for human-robot interaction, but they also harbor risks such as misinterpretation, bias, and safety issues. The University of Augsburg's subproject systematically investigates the risks associated with the use of LLMs and LC/NC mechanisms based on them in nursing care.

The aim is to advance methods for early risk detection and optimization of language models and to promote their application in the Module 1 projects. To raise awareness among nursing staff, practical test scenarios are being developed, tested, and evaluated in a real-world laboratory. The insights and methods gained contribute to the improvement of existing systems and support the safe, transparent use of LLMs in healthcare robotics.