Trust by Design: Building Reliable AI for Sensitive Environments

In a world increasingly shaped by AI-driven decisions, trust isn't just a value – it's a requirement for all stakeholders who are impacted by these decisions. From legal tech and healthcare to public administration, AI assistants are entering domains where mistakes aren’t just inconvenient – they're consequential. The keynote of Dr. Kim Nguyen and Prof. Dr. Gesche Joost explores what it truly means to design trustworthy AI with a human-in-the-loop approach for critical or sensitive areas. What are the key dimensions of trust in AI systems? How can we ensure correctness, traceability, data security and sovereignty? And how do we evaluate whether a Large Language Model (LLM) is truly “fit for purpose”? Benchmarking and continuous model evaluation will play a central role in this discussion. As well as finding methods to evaluate and contextualize trust based the motivations and behaviors between users and AI tools such as LLMs.

Drawing on current research from Bundesdruckerei and the German Research Center for Artificial Intelligence (DFKI), this session provides concrete insights, guiding questions, and practical approaches to help organizations move from mere AI adoption to intentional, responsible integration.