Position: Modular Safety Guardrails Are Necessary for Foundation-Model-Enabled Robots in the Real World

Joonkyung Kim, Wenxi Chen, Davood Soleymanzadeh, Yi Ding, Xiangbo Gao, Zhengzhong Tu, Ruqi Zhang, Fan Fei, Sushant Veer, Yiwei Lyu, Minghui Zheng, Yan Gu

TL;DR: We argue that foundation-model-enabled robot safety requires safety notions beyond physical constraint satisfaction and advocate modular safety guardrails for comprehensive safety in real-world robotic deployment.

Abstract: The integration of foundation models (FMs) into robotics has accelerated real-world deployment, while introducing new safety challenges arising from open-ended semantic reasoning and embodied physical action. These challenges require safety notions beyond physical constraint satisfaction. In this position paper, we characterize FM-enabled robot safety along three dimensions: action safety (physical feasibility and constraint compliance), decision safety (semantic and contextual appropriateness), and human-centered safety (conformance to human intent, norms, and expectations). We argue that existing approaches, including static verification, monolithic controllers, and end-to-end learned policies, are insufficient in settings where tasks, environments, and human expectations are open-ended, long-tailed, and subject to adaptation over time. To address this gap, we propose modular safety guardrails, consisting of monitoring (evaluation) and intervention layers, as an architectural foundation for comprehensive safety across the autonomy stack. Beyond modularity, we highlight possible cross-layer co-design opportunities through representation alignment and conservatism allocation to enable faster, less conservative, and more effective safety enforcement. We call on the community to explore richer guardrail modules and principled co-design strategies to advance safe real-world physical AI deployment.

Lay Summary: As foundation models become increasingly integrated into robotics, robots are gaining the ability to perform more complex tasks in real-world environments. However, these advances also introduce new safety challenges that go beyond simply avoiding physical collisions or satisfying predefined constraints. Robots must also make appropriate decisions in open-ended situations and interact safely with humans while respecting human expectations and intentions. In this position paper, we describe robot safety from three perspectives: safe physical actions, safe decision-making, and human-centered behavior. We argue that existing safety approaches are often not sufficient for real-world settings where environments, tasks, and human preferences can change over time and may be difficult to predict in advance. To address these challenges, we propose the use of modular safety guardrails that monitor robot behavior and intervene when unsafe situations are detected. We further discuss how different safety components across the robotic system can be designed to work together more effectively. Finally, we encourage the robotics community to develop more advanced safety mechanisms and system-level design strategies to support the safe deployment of future physical AI systems.

Primary Area: System Risks, Safety, and Government Policy

Keywords: Foundational model, Robotics, Embodied AI, Safety

Submission Number: 694