Study Finds Linguistic Style Can Serve as Hidden Attack Surface for LLM Safety
A new arXiv paper argues that linguistic style is a systematic and previously overlooked attack surface, showing that varied styles can jailbreak LLMs across a diverse taxonomy of styles and multiple models.
Read the full story at arXiv cs.CL ↗
Timeline · 1 report
- 2026-11-17 05:00 · arXiv cs.CL
Jailbreak LLMs with Linguistic Style as a Hidden Attack Surface