Our paper “A Unified Safety Subspace Exists in Speech Language Models” has been accepted at Interspeech 2026! 🎉