Year: 2026
Type: conference-paper
Abstract: Safety alignment of large language models (LLMs) is mostly evaluated in English and contract-bound, leaving multilingual vulnerabilities understudied.We introduce Indic Jailbreak Robustness (IJR), a judge-free benchmark for adversarial safety across 12 Indic and South Asian languages ( 2.1 Billion speakers), covering 45,216 prompts in JSON (contractbound) and FREE (naturalistic) tracks.IJR reveals three patterns.(1) Contracts inflate refusals but do not stop jailbreaks: in JSON, LLaMA and Sarvam exceed 0.92 JSR, and in FREE all models reach ≈1.0 with refusals collapsing.(2) English→Indic attacks transfer strongly, with format wrappers often outperforming instruction wrappers.(3) Orthography matters: romanized/mixed inputs reduce JSR under JSON, with correlations to romanization share and tokenization (ρ ≈ 0.28-0.32)indicating systematic effects.Human audits confirm detector reliability, and lite-to-full comparisons preserve conclusions.IJR offers a reproducible multilingual stress test revealing risks hidden by English-only, contract-focused evaluations, especially for South Asian users who frequently code-switch and romanize.Warning: This paper may contain offensive or harmful content, included solely for scientific transparency.
Authors: Priyaranjan Pattnayak, Sanchari Chowdhuri
Institution: Oracle (United States)
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7140140875