When Robots Mishear Us: Mapping the Safety Risks of Voice-Controlled Embodied AI
2026-08-28 • Artificial Intelligence
Artificial IntelligenceComputation and LanguageRobotics
AI summaryⓘ
The authors studied if mistakes made by speech-to-text systems (ASR errors) can cause AI robots to do unsafe things. They tested these errors using special safety tests and found that some mistakes make instructions confusing or unclear, while others stop the AI from refusing dangerous tasks. They also tried fixing these errors automatically, but it didn't always help. Overall, the authors found that speech recognition mistakes can create real safety problems for AI that acts in the physical world.
Embodied AIAutomatic Speech Recognition (ASR)Safety BenchmarksSafeAgentBenchPOEXModel Refusal BehaviorHarmful AmbiguityASR Error CorrectionUnsafe OutputsAI Safety
Authors
Sihan Jia, Oliver Lemon
Abstract
We investigate whether automatic speech recognition (ASR) errors in user input can lead to unsafe outputs from Embodied AI (EAI) models. We find that ASR errors can lead to harmful instructions being accepted and executed by EAI models, thereby reducing safety. We simulate ASR errors and combine them with existing safety benchmarks (SafeAgentBench and POEX) to evaluate how different errors affect embodied AI safety. We find that some of them preserve semantic structure but increase harmful ambiguity, while others weaken the model refusal behaviour and allow unsafe plans to be generated and executed. We show that in some cases automatic correction of ASR errors can reduce the risk, but this is not always effective. Overall, we show that ASR errors lead to significant safety risks for embodied AI.