Comparing Notes on Trustworthy AI, 18.05.

Embodied Intelligence: The Physical AI Moment

Something fundamental is shifting in AI.
Systems are no longer confined to screens and servers; they are acquiring bodies, building models of the physical world and learning to act within it. Humanoid robots, embodied AI and world models represent a new frontier that raises questions software AI never had to answer: What happens when an AI system can anticipate, plan and intervene in shared physical space?

This opening panel of the Comparing Notes on Trustworthy AI event series brings together a human-robot interaction researcher studying trust in embodied systems, a scientist building assistive AI for people with severe motor impairments and an ethicist working on human rights frameworks for AI governance to take stock of where Physical AI actually stands in 2026 and what responsible development demands of us.

Panelists:

• Olivia Herzog is a Science Transfer expert at robo.innovate (Munich Institute of Robotics and Machine Intelligence), TUM Venture Labs Robotics/AI, and the Robotics Institute Germany (RIG). Bridging the gap between cutting-edge research and the market, she leverages her background in Human-Robot Interaction (HRI) and psychology to architect the next generation of European robotics ventures. Her doctoral research at TUM focuses on closing the "trust gap" in Physical AI, specifically investigating how mobile robots navigate public spaces and sensitive settings. By exploring the social cues that drive human-robot collaboration, Olivia provides the empirical foundation for the responsible, trustworthy deployment of embodied systems in shared human environments.
• Annette Hagengruber is a research scientist and Team Lead at the German Aerospace Center (DLR), where she leads the team re-enabling robotics. Her work focuses on human-centered robotics, combining artificial intelligence and real-world system integration to develop assistive technologies for people with severe motor impairments. Her research has been published in leading high-impact journals and conferences. Their approach and system were validated under real-world conditions by winning the CYBATHLON 2024 in the category Assistance Robot Race, an international competition organized by ETH Zürich. Besides her role at DLR, she is co-founding Enabling Robotics, a technology transfer initiative translating their research into a scalable robotics company. The startup develops AI-based robotic manipulation systems that enable real-world deployment and continuous improvement through a human-in-the-loop approach. Starting with assistive robotics for wheelchair users, where the user is an intrinsic part of the system, the company generates real-world data and validated autonomy functions to scale into broader service robotics applications. The long-term vision is to enable robots to reliably perform physical tasks in real world environments.
• Dr. Alexander Kriebitz is a political scientist and ethicist specializing in international law, business ethics, and AI ethics. He is a postdoctoral researcher at the Technical University of Munich and Ludwig Maximilian University of Munich, where he teaches on anthropology and moral philosophy. His work focuses on integrating human rights into the ethical development and governance of artificial intelligence. Alexander has published in journals such as the Business and Human Rights Journal, AI and Ethics, and Human Rights Review.

About the series

• This event is part of the "Comparing Notes on Trustworthy AI" series.
• Hosted by: appliedAI Institute for Europe in collaboration with LMU Munich Centre for ML, Liquid Legal Institute e.V., and TUM Think Tank.
• This project is funded by the Bavarian Ministry for Digital Affairs.

The detailed agenda and further event information will be shared soon..