SOURCE-LINKED INTELLIGENCE

When Safety Speaks a Language: A Mechanistic Analysis of Safety-Language Identity Entanglement in LLMs

arXiv · AI, language, vision and robotics · article · Aug 30, 2026 · UTC

Safety alignment of large language models (LLMs) degrades across languages, yet the internal mechanism driving this asymmetry remains poorly understood. Our work, therefore, presents a systematic mechanistic analysis of multilingual safety using sparse autoencoder (SAE) features, sparse interpretable directions in the residual stream associated with harmful and harmless model behavior across three instruction-tuned LLMs, eight languages, and all model layers. We observe that safety-relevant features are architecture-dependent in terms of where they are located and how they are distributed acro

Read original source ↗ Open in workspace

recordType

paper

region

Global